The Art and Science of Low Carbohydrate Living: An Expert Guide to Making the Life-Saving Benefits of Carbohydrate Restriction Sustainable and Enjoyable

Overall score

Scientific accuracy

Reference accuracy

Healthfulness

How hard would it be to apply the book's advice? Fairly difficult

Book published in 2011

Published by Beyond Obesity LLC

First Edition, Paperback

Review posted August 28, 2023

Primary reviewer: Stephan Guyenet

Peer reviewer: Mario Kratz

Table of contents

1. Introduction
2. Scientific accuracy
3. Reference accuracy
4. Healthfulness
5. Most unusual claim
6. Other
7. Conclusion
8. Updates

Claim 1

Low-carbohydrate diets increase insulin sensitivity (reduce insulin resistance)

Supporting quote(s) and page number(s)

Back cover: “Carbohydrate restriction is the proverbial ‘silver bullet’ for managing insulin resistance, metabolic syndrome and type-2 diabetes.”

Page 76: “The primary reason we have an entire chapter about insulin resistance is that well-formulated low carbohydrate diets consistently make it better.”

Criterion 1.1. How well is the claim supported by current evidence?

2.5 out of 4

This claim received a score of 2.5, meaning it’s weakly to moderately supported by current evidence. The reason the score is 2.5 rather than an integer is that we scored two versions of the claim and averaged those scores together. We scored two versions of the claim because it’s not clear which specific claim Art and Science is making.

Before proceeding, it’s important to note that Art and Science is referring to a certain range of carbohydrate intakes when it talks about the benefits of low-carbohydrate diets for insulin resistance. The section “What Does ‘Low Carbohydrate’ Mean” (page 3) discusses what that range is, and mentions 125 grams of carbohydrate per day at the upper end, and 20 grams per day at the lower end. Assuming an average 2,500 calorie diet, that works out to about 3% to 20% of total calories. This is in contrast to the typical US intake of about 47% of calories (see “Nutrient availability” estimates from the USDA Economic Research Service, 2010).

The book’s claim is hard to evaluate because it’s not entirely clear if it’s arguing that low-carbohydrate diets increase insulin sensitivity due to weight loss and possibly other factors, or independently of weight loss. This matters because if Art and Science is arguing that low-carb diets tend to cause weight loss in people who carry extra fat, and weight loss tends to improve insulin sensitivity, that’s a straightforward claim that’s easy to support. If it’s arguing that low-carb diets increase insulin sensitivity whether or not a person loses weight, that’s a bolder claim that’s harder to support.

Art and Science certainly doesn’t emphasize weight loss as an explanation for increased insulin sensitivity, focusing on other explanations like the possibility that reduced insulin exposure may allow the insulin response to resensitize (p. 186-7). On page 188, it argues that the metabolic benefits of carbohydrate restriction are independent of weight loss, but doesn’t specifically mention insulin sensitivity: “The benefits of carbohydrate restriction are also apparent whether weight loss occurs or not, suggesting it is the reduction in dietary carbohydrate, not calories or body fat mass per se, that is responsible for the improved metabolic outcomes.”

We aren’t sure which version of the argument the book is making so we chose to split it down the middle. We evaluated each version in turn, and averaged them.

The first version of the claim is that low-carbohydrate diets cause weight loss, and weight loss increases insulin sensitivity. A 2020 network meta-analysis (study of studies) of weight loss trials reports that at 6 months, low-carbohydrate diets cause an average of 10 lbs (4.6 kg) of weight loss in people with obesity or overweight. In the same analysis, the Atkins diet, which is a classic and popular version of a low-carbohydrate diet, caused an average of 12 lbs (5.5 kg) of weight loss. The argument that low-carbohydrate diets cause weight loss in people with excess body fat is well supported.

To understand the impact of weight loss on insulin sensitivity, we focus on studies that used the gold-standard method of measuring insulin sensitivity: the euglycemic-hyperinsulinemic clamp (more on why this is important later). A number of studies have reported that weight loss increases insulin sensitivity when it’s measured in this way. This is especially obvious with major weight loss, for example caused by bariatric surgery. Since low-carbohydrate diets cause weight loss, and weight loss increases insulin sensitivity, we think this version of the claim is well supported. This corresponds to a score of 4 out of 4.

How about the second version of the claim, that low-carb diets increase insulin sensitivity independently of weight loss? Here, weight loss is out of the picture, and we have to look for studies that either reduced carbohydrate intake without causing weight loss, or compared a low-carbohydrate diet to a higher-carbohydrate diet that both caused about the same amount of weight loss.

[Warning: the following two paragraphs are unavoidably technical] This narrows the field quite a bit, and we have to narrow it further by looking for studies that measured insulin sensitivity using the gold-standard euglycemic-hyperinsulinemic clamp, or other direct measures of insulin sensitivity. The reason this is important is that indirect measures of insulin sensitivity, like the commonly-used HOMA method, may be misleading in the context of a low-carb diet. This is because HOMA and related methods are based in part on fasting insulin level, and fasting insulin probably doesn’t just depend on insulin sensitivity, but also on insulin demand. Because there’s less glucose flux to deal with in the post-meal and fasting states on a low-carb diet, insulin demand and insulin secretion are lower, but this doesn’t necessarily mean insulin sensitivity is higher.

HOMA was designed to be an easy measure of insulin sensitivity that correlates reasonably well with the euglycemic-hyperinsulinemic clamp in regular people eating regular diets, but to our knowledge it hasn’t been validated as an accurate measure of changes in insulin sensitivity with carbohydrate restriction. Most, but not all, of the evidence cited in Art and Science to support its claims about insulin sensitivity is based on HOMA (more on this later).

We performed a scientific literature search to look for studies that studied the impact of low-carbohydrate diets on insulin sensitivity independent of weight loss, using direct measures of insulin sensitivity (particularly the euglycemic-hyperinsulinemic clamp). Our search included the following:

1. A PubMed keyword search. The search terms were “carbohydrate AND diet AND clamp AND insulin AND (sensitivity OR resistance)” and the search was performed on May 8, 2023. The “clinical trial” box was checked. This returned 275 results, all of which we considered for relevance.
2. We looked for additional studies via the “similar articles” feature in PubMed.
3. We looked for additional studies via the references in the studies we identified.
4. We considered the references cited in Art and Science.

Studies had to include at least one group with 20 percent carbohydrate intake or less, and they had to either not cause weight loss, or compare a low-carbohydrate diet to a higher-carbohydrate diet that both caused about the same amount of weight loss.

This search returned 6 studies:

1. Bradley 2009 “compared a low-fat (20% fat, 60% carbohydrate) versus a low-carbohydrate (60% fat, 20% carbohydrate) weight reduction diet in 24 overweight/obese subjects… in an 8-week randomized controlled trial.” Both diets caused about the same amount of weight loss. The study measured insulin sensitivity using euglycemic-hyperinsulinemic clamp and reports that low-carbohydrate and low-fat diets cause “comparable effects on insulin resistance”.
2. Kirk 2009 compared calorie-restricted diets that were either low-fat (20% fat, 65% carbohydrate) or low-carbohydrate (75% fat, 10% carbohydrate), each for 11 weeks, in 22 people with obesity. Both diets caused about the same amount of weight loss. The study measured insulin sensitivity using euglycemic-hyperinsulinemic clamp and reports no significant difference between diets.
3. Chokkalingam 2007 compared diets that were either typical (35% fat, 50% carbohydrate) or low-carbohydrate (75% fat, 10% carbohydrate), each for 6 days. Diets were eaten in random order by 10 “healthy men”. The study reports that insulin sensitivity as measured by hyperinsulinemic-euglycemic clamp (Rd) was “slightly elevated” (statistically significant) by 10% in the low-carbohydrate diet condition relative to the typical diet condition.
4. Bisschop 2001 compared diets that were either very-high-carbohydrate (0% fat, 85% carbohydrate), typical (41% fat, 44% carbohydrate), or very-low-carbohydrate (83% fat, 2% carbohydrate), each for 11 days. Diets were eaten in random order by 6 “healthy men”. The study reports that whole-body insulin sensitivity as measured by hyperinsulinemic-euglycemic clamp was the same between diets, but liver insulin sensitivity was lowest on the very-low-carbohydrate diet.
5. Cutler 1995 compared diets that were either typical (35% fat, 51% carbohydrate) or low-carbohydrate (75% fat, 8% carbohydrate) for three weeks. Participants were ten exercise-trained “healthy men”. The study reports that whole-body insulin sensitivity as measured by hyperinsulinemic-euglycemic clamp was the same between diets.

We found one additional study, Himsworth 1935, but it was published nearly a century ago and we have not been able to get the full text. We’re excluding it because we don’t have enough information to critically evaluate it, and in any case its methods probably aren’t up to modern standards.

The ideal study for this question would be one that compares a very-low-carbohydrate diet to a higher-carbohydrate diet for 3 or more weeks in people who are insulin resistant, with no weight change in either diet group, and compares insulin sensitivity between groups at the end using hyperinsulinemic-euglycemic clamp. None of the studies we found correspond to this ideal. Some involved weight loss in both diet groups, which could mask possible benefits of a low-carbohydrate diet, and some involved participants who were likely insulin-sensitive at baseline. Some of the studies were very small or very short. So the findings have to be interpreted with a lot of caution.

That said, the weight of this evidence suggests that there is little or no difference in whole-body insulin sensitivity between very-low-carbohydrate and higher-carbohydrate diets when weight loss is controlled for and insulin sensitivity is directly measured. Given the uncertainty involved, this provisional conclusion could be easily overturned by more research.

Art and Science cites three studies to support its claim that low-carb diets increase insulin sensitivity. None of these met the criteria for our scientific literature search, but we’ll discuss them here. One study (cited on p. 87) reports that in people with type 2 diabetes, a 14-day low-carb diet increased insulin sensitivity as measured by hyperinsulinemic-euglycemic clamp. However, the diet caused weight loss so the study doesn’t tell us whether insulin sensitivity changed independently of weight loss.

A second study (companion paper to the one cited on p. 87, which doesn’t report on insulin sensitivity) reports that a 12-week low-calorie, low-carb diet causes an apparent increase in insulin sensitivity relative to a low-calorie, low-fat diet. However, this study has two limitations for our current purposes: the low-carb diet caused more weight/fat loss than the low-fat diet, and they measured insulin sensitivity using the HOMA method, whose accuracy is unclear in this context. Volek and Phinney are authors on the paper.

A third study is cited on p. 84, but its findings on insulin sensitivity are not mentioned in the book. Eight weight-stable men with obesity or overweight were given two 6-week low-carb diets, one high in saturated fat and the other high in unsaturated fat. The diets were designed to prevent weight loss. The study used HOMA as a measure of insulin sensitivity. It reports that “Blood glucose, insulin, and HOMA-IR were not significantly different from baseline or between diets.” To be fair, the number of participants may have been too small to detect an effect, but on the other hand this study seems more relevant than the other two cited in Art and Science to address the claim that low-carb diets increase insulin sensitivity independent of weight loss. Volek and Phinney are authors on the paper.

The second version of the claim, that a low-carbohydrate diet increases insulin sensitivity independent of weight loss, is poorly supported by current evidence. This corresponds to a score of 1 out of 4. Averaging this together with the previous score of 4, this yields a scientific accuracy score of 2.5 out of 4 for this claim.

Overall (average) score for claim 1

2.5 out of 4

Claim 2

Low-carbohydrate diets are a particularly effective way of managing type 2 diabetes

Supporting quote(s) and page number(s)

Back cover: “Carbohydrate restriction is the proverbial ‘silver bullet’ for managing insulin resistance, metabolic syndrome and type-2 diabetes.”

Page 191: “If this is indeed a primary underlying pathophysiology of type-2 diabetes, then it follows that the optimum treatment of type 2 diabetes is reduced dietary carbohydrate intake.”

Page 192: “…a well formulated low carbohydrate diet will tend to produce striking improvements when implemented in type-2 diabetics.”

Criterion 1.1. How well is the claim supported by current evidence?

3 out of 4

This claim received a score of 3 out of 4, meaning that it’s moderately supported by current evidence. Low-carb diets appear to be among the best diets for managing type 2 diabetes, and they can be highly effective in motivated people with strong support, but in average people receiving average support they aren’t highly effective or clearly superior to other diets.

Type 2 diabetes is a disease in which insulin isn’t able to do its job very well, usually because the body isn’t responding to it very well (insulin resistance) and the pancreas isn’t able to secrete enough of it (beta cell dysfunction). The disease is diagnosed by high blood sugar, either in the fasting state, by assessing the average blood sugar level over about 3 months using the HbA1c test, or after drinking a beverage with a standard amount of sugar in it (oral glucose tolerance test).

It makes sense that low-carbohydrate diets would have a special advantage in managing type 2 diabetes, since they reduce the amount of insulin that’s needed for metabolic control (insulin demand). This is because there is less carbohydrate entering the body from the diet, so less insulin has to be secreted to cover it. This would be expected to result in better metabolic control including better blood sugar levels, and a need for fewer medications like injected insulin. The weight loss produced by low-carbohydrate diets should help too, although it’s not clear that they have a major advantage over other diets in that regard.

That’s the theory, but what’s the evidence? Lukas Schwingshackl and colleagues published a network meta-analysis (study of studies) in 2018 that addresses this question. Only randomized controlled trials, the most rigorous type of diet trial, were included. They placed diets into 8 categories: low-carb, Mediterranean, Paleolithic, vegetarian, low-glycemic, moderate-carb, high-protein, and low-fat. They determined the impact of these diets on HbA1c, a marker of overall blood sugar control, and fasting blood glucose.

All 8 diets significantly reduce HbA1c and fasting glucose to some degree, relative to eating as usual. Of these 8 diets, low-carb is the most effective at reducing HbA1c, although the Mediterranean diet is close behind. The Mediterranean diet is the most effective at reducing fasting glucose. The findings could be interpreted in different ways, but in our opinion they suggest that low-carb diets are more effective than most diets at improving blood glucose control, although not necessarily more effective than the Mediterranean diet.

It’s important to note that in real-world trials like these, people don’t usually stick to the diet very well, especially as the months turn into years. We think these trials are informative because they tell us what happens when people actually try to incorporate these diets into their lives in the real world.

That said, what would happen if the diet were tried in highly motivated people given lots of support so they could stick with it better? We think a good example of this is the Virta Health trial, which Volek and Phinney were involved in. Volek and Phinney are co-founders of Virta Health, an online medical clinic and coaching service that emphasizes low-carbohydrate nutrition for managing type 2 diabetes.

The study wasn’t a randomized controlled trial in which people are randomly assigned to one diet or another. Instead, it was a non-randomized controlled trial in which people with type 2 diabetes got to choose whether they wanted to be enrolled in the Virta Health intervention, or receive standard diabetes care. This means that the people in the Virta group were enthusiastic about trying a low-carb diet, and received extensive (and expensive) support relating to diet, lifestyle, and medications. It also means the results can’t be compared to the randomized controlled trials we discussed previously, which weren’t conducted in such favorable conditions.

The Virta research team has published results for 1- and 2-year follow-up in people receiving the intervention.

• The 1-year follow-up reports on 349 adults with type 2 diabetes. One year after the start of the study, the results show a large reduction in HbA1c (60 to 45 mmol/mol), a large reduction in body weight (-30 lbs; 14 kg), and a reduction in the percentage of people who needed diabetes medications other than metformin (from 57% at baseline to 30% at endline).
• The 2-year follow-up reports that these improvements were mostly maintained over two years, although they did deteriorate a bit. Keep in mind that this only includes people who stayed enrolled in the program, which likely exaggerates its benefits because the people who experience fewer benefits are more likely to drop out and not be included in the data.

We think these results suggest that a low-carbohydrate diet can be quite effective as part of a broader program for managing type 2 diabetes in motivated and well-supported people.

That said, the scientific literature as a whole doesn’t quite support the contention in Art and Science that the diet is a “silver bullet” for managing type 2 diabetes. In the general population, with typical levels of motivation and support, low-carbohydrate diets are a good choice for managing type 2 diabetes but are not highly effective or markedly superior to other diets.

Overall (average) score for claim 2

3 out of 4

Claim 3

A higher intake of saturated fat is not harmful in the context of a low-carbohydrate diet

Supporting quote(s) and page number(s)

Back cover: “Dietary saturated fat is not a demon when you are low carb adapted.”

Page 41: “If it doesn’t accumulate when you eat it, and eating it hasn’t been shown to actually harm you, where’s the demon in saturated fat?”

Page 100: “It turns out most of the harmful effects attributed to dietary saturated fat (e.g., increased heart disease, insulin resistance, vascular dysfunction, etc.) are unsubstantiated. The truth is that saturated fats only become a problem when they accumulate. And the guilty party for saturated fat accumulation, in most cases, is dietary carbohydrate.”

Page 108: “But clearly, in the context of a keto-adapted individual following a low-carbohydrate diet, where saturated fat disposal is accelerated causing blood levels to drop, there is no basis to be concerned about their inclusion in the foods we choose to eat.”

Criterion 1.1. How well is the claim supported by current evidence?

2 out of 4

This claim received a score of 2 out of 4, meaning it’s weakly supported by current evidence. The claim can be interpreted in two ways, one of which we think is fairly well supported (3 out of 4) and the other of which is poorly supported (1 out of 4). We averaged these scores together to yield the overall score for this claim.

The primary concern with saturated fat is its impact on cardiovascular risk via the concentration of LDL-cholesterol in the blood (or the concentration of ApoB protein that is associated with it). Low-carb diets don’t have to be high in saturated fat. This claim can be interpreted in two ways:

1. A person can eat a low-carb diet high in saturated fat and not experience an overall increase in cardiovascular risk, relative to typical diets.
2. A low-carb diet high in saturated fat is just as heart-healthy as a low-carb diet low in saturated fat.

To address the first angle, we sought meta-analyses of randomized controlled trials reporting the impact of low-carb diets on markers of cardiovascular risk. Due to the abundance of options, we focused on network meta-analyses published in the last 5 years (2018-2023). We didn’t consider meta-analyses that only compared low-carbohydrate diets against a specific other type of diet, like a low-fat diet.

We identified one recent network meta-analysis that includes trials regardless of the participants’ disease status, and one that only includes people with type 2 diabetes.

• Ge 2020 is a network meta-analysis of randomized controlled trials that compares the impact of 14 popular diets (including all low-carb diets pooled together, and Atkins separately) on weight loss and cardiovascular risk factors in people with obesity/overweight, regardless of disease status. It reports that at 6 months, compared to usual diet, low-carbohydrate diets reduce body weight by 10 lbs (4.6 kg), systolic blood pressure by 5 mmHg, and do not significantly impact LDL-cholesterol. Also at 6 months and compared to usual diet, the Atkins diet reduces body weight by 12 lbs (5.5 kg), systolic blood pressure by 5 mmHg, and slightly increases LDL-cholesterol and HDL-cholesterol (~3 mg/dL). Effects on body weight and cardiovascular risk factors were smaller at 12 months.
• Bonekamp 2023 is a network meta-analysis of randomized controlled trials that compares the impact of 8 dietary patterns on weight loss and cardiovascular risk factors in people with type 2 diabetes. It reports that at 6 months, low-carb diets reduce body weight by 11 lbs (4.8 kg), reduce HbA1c by 0.8 mmol/mol (indicating better blood sugar control), and do not significantly impact blood pressure or blood lipids. Changes in body weight (-1.3 kg) and HbA1c (+0.1 mmol/mol) were no longer significant at 12 months.

This suggests that low-carb diets tend to have a favorable overall effect on cardiovascular risk profile in the context of weight loss, despite the fact that they are usually higher in saturated fat. This supports the first interpretation of the book’s claim that saturated fat isn’t harmful in the context of a low-carb diet, but keep in mind that these are just biomarkers of cardiovascular risk, not data on actual cardiovascular event rates. So while it looks like low-carb diets are probably good for the heart if we factor in the weight loss they commonly cause, we don’t have strong evidence of it.

One specific concern, for example, is that the overall improvements seen in cardiovascular risk factors may only be present while people are actively losing weight, i.e., these improvements may be temporary. We assign this version of the claim a 3 out of 4, meaning it’s moderately well supported by evidence.

It’s also worth mentioning that the individual response to any diet can vary so it’s still a good idea to monitor blood lipids for unwanted changes.

Yet low-carbohydrate diets don’t have to be high in saturated fat, and a low-carbohydrate diet high in saturated fat could still be suboptimal. To address the second interpretation of the book’s claim, we sought studies that compare low-carbohydrate diets high in saturated vs. unsaturated fat. We searched PubMed with the search terms “low-carbohydrate AND diet AND saturated AND unsaturated” on June 5th, 2023.

This returned one randomized controlled trial, by Volek and Phinney. This is the only trial we know of that compares low-carbohydrate diets high vs. low in saturated fat. Eight weight-stable men with obesity or overweight were fed a low-carbohydrate diet high or low in saturated fat, in random order, for 6 weeks (with 7 weeks of washout between diets).

Both diets provided 12-13% of calories as carbohydrate, 58-59% as fat, and 29-30% as protein. The high-saturated-fat diet provided 31% of calories as saturated fat, while the low-saturated-fat diet provided 17% of calories as saturated fat. It’s worth noting that the low-saturated-fat diet still contained a fair amount of saturated fat, both in absolute terms and as a percentage of total fat (~29%). All food was provided, which increases our confidence that participants ate what the study intended.

The paper, and Art and Science, focuses on changes in the types of fats circulating in the blood, concluding that saturated fat intake has little impact on levels of saturated fat in the blood. While this is an interesting and surprising finding, its relevance to cardiovascular risk is unclear.

What we’re more interested in is the impact of the diets on blood lipids linked to cardiovascular risk. Comparing between diets, there were no significant differences in LDL-cholesterol, HDL-cholesterol, or triglycerides. That said, even though it wasn’t statistically significant (p=0.119), LDL-cholesterol was 15% higher on the high-saturated-fat diet, which is a clinically significant difference if it’s real. Given that the study only included 8 people, there’s a good chance that it was simply too small to detect a difference. Given that diets high in saturated fat tend to increase LDL-cholesterol in people eating typical diets, we would expect this to probably also be the case on low-carb diets, and these findings don’t refute that.

We don’t think this study supports the book’s contention that saturated fat isn’t harmful in the context of a low-carb diet. There was a trend toward higher LDL-cholesterol on the high-saturated-fat diet, and the study was likely too small to tell whether this difference was real.

We also found a second study, which is a nonrandomized secondary analysis of the DIETFITS randomized trial. Although the original DIETFITS trial was one of the most rigorous diet trails comparing low-fat vs. low-carb diets, this analysis is much less informative because it compared people within groups rather than between groups. This makes it more of an observational study than a controlled trial.

The study reports that among people on the low-carb diet, changes in saturated fat intake over the 12-month duration of the trial were weakly correlated with changes in LDL-cholesterol, and this correlation was only statistically significant in one out of two analyses. The correlation was a bit stronger in people eating a low-fat diet, and it was statistically significant in both analyses. We don’t think this study provides much evidence either way.

We assign this version of the claim a 1 out of 4, meaning the evidence doesn’t support the book’s claim but doesn’t refute it either.

 

Overall (average) score for claim 3

2 out of 4

Overall (average) score for scientific accuracy

2.5 out of 4

Reference 1

Reference

Chapter 14, reference 107.  Sonsken and Sonsken. Br J Anaesth 85:69. 2000

Associated quote(s) and page number(s)

Page 179: “An equally compelling case can be made for defects in hepatic insulin action as the driving force underlying high blood glucose.” The passage goes on to describe the “open faucet theory”, in which high blood sugar in insulin resistance is caused by a reduced ability of insulin to suppress glucose release by the liver into the bloodstream. Importantly, this idea is mentioned as one of two possible explanations for high blood sugar in people with insulin resistance, the other being the “clogged drain theory”, in which high blood sugar in insulin resistance is caused by a reduced ability of insulin to stimulate glucose uptake from the bloodstream.

Criterion 2.1. Does the reference support the claim?

4 out of 4

This reference received a score of 4 out of 4, meaning that it strongly supports the passage in Art and Science. The cited paper is mostly about insulin and glucose function in the context of type 2 diabetes, and it doesn’t mention metabolic syndrome, the topic of chapter 14 of Art and Science. It does briefly discuss prediabetes, which is similar (and they both involve insulin resistance).

The paper argues strongly in favor of the “open faucet theory” and against the “clogged drain theory”, although it doesn’t use those terms.

This one is arguable but we think a score of 4 out of 4 is the best fit here, for these reasons:

1. Although it’s mostly in the context of diabetes, the paper does discuss the open faucet and clogged drain theory in a way that’s consistent with the passage in Art and Science.
2. Art and Science doesn’t argue for one theory over the other, it just explains them both as possibilities. The book isn’t actually making a claim about which one explains high blood sugar in insulin resistance, so it would be hard for a reader to be misled.

Reference 2

Reference

Chapter 10, reference 86.  Bergström et al. Acta Physiol Scand 71:140. 1967

Associated quote(s) and page number(s)

Page 128: “Starting with the classic studies of Christensen and Hansen before World War II, a string of short-term studies have been published demonstrating longer endurance exercise times with high carbohydrate diets compared to low carbohydrate diets.”

Criterion 2.1. Does the reference support the claim?

4 out of 4

This reference received a score of 4 out of 4, meaning that it strongly supports the passage in the book. It’s a diet trial published in 1967 reporting that a high-carbohydrate diet allows people to bicycle at 75% of maximum capacity for longer than a high-fat diet. The diet periods were only a few days long, consistent with the definition Art and Science uses for a “short-term” study: less than 14 days.

The paper cites a previous trial with similar results, published in 1939 (“Christensen and Hansen” in the passage above). Although it’s debatable whether two studies constitute a “string”, this seems like too small of a problem to dock a point.

Reference 3

Reference

Chapter 7, reference 32.  Kannel, Anderson, and Wilson. JAMA 267:1253. 1992

Associated quote(s) and page number(s)

Page 81: “About 20 years ago, it was noted that people with persistently high biomarkers of inflammation (e.g., CRP and IL-6) were at increased risk of heart attack.”

Criterion 2.1. Does the reference support the claim?

4 out of 4

This reference received a score of 4 out of 4, meaning that it strongly supports the passage in the book.

The reference reports the relationship between white blood cell count and cardiovascular disease risk in about 2,800 people (Framingham cohort). White blood cell count is a marker of inflammation. The paper concludes: “The degree of elevation of WBC count within the normal range is a marker for increased risk of [cardiovascular disease] that is partially explained by cigarette smoking.”

This study was published in 1992, consistent with the quote’s claim that this relationship was first seen “about 20 years ago”. Art and Science was published in 2011, 19 years after this paper.

The other study Art and Science cites to support the passage reports a relationship between the biomarker of inflammation C-reactive protein (CRP) and heart attack risk. This is helpful because the passage mentions CRP and IL-6 but not white blood cell count specifically.

Reference 4

Reference

Chapter 18, reference 133.  Heymsfield et al. Metabolism 38:215. 1989

Associated quote(s) and page number(s)

Page 242: “It takes about 350 miles of running or 1,000 miles of cycling to burn off about 10 pounds of body fat (assuming that your appetite doesn’t increase or your metabolism slows down). Unfortunately, when heavy people exercise regularly, their resting metabolism slows– this is not a typo– it SLOWS by 5 to 15% on average. Based on the results of 4 tightly controlled, inpatient human studies, instead of losing 10 pounds, the average person loses 7 pounds with this much exercise, and some people lose as little as 2 or 3.” Reference 133 appears at the end of this quote, along with three other studies.

Criterion 2.1. Does the reference support the claim?

2 out of 4

This reference received a score of 2 out of 4, meaning that it weakly supports the passage in the book.

The reference is a study in which 11 women with obesity were placed on a very-low-calorie diet (900 kcal/day), with or without exercise, for five weeks. The exercise was 346 kcal/day of aerobic activity. As stated in Art and Science, it was an inpatient study, which means the participants stayed in a research facility for the duration of the study.

The first thing to note here is that the calorie deficit in this study was mostly due to the diet rather than the exercise. According to the paper, the women ate about 2,100 kcal/day at baseline, which implies their diet was causing a calorie deficit of 1,200 kcal/day. This is much larger than the 346 kcal/day deficit caused by the exercise. This is an important feature of this study that isn’t mentioned in Art and Science. It’s also worth noting that this was a very small study, with only 5-6 people in each group, so its findings should be taken with a grain of salt.

The diet and exercise group lost 12 lbs (5.3 kg) of body fat, while the diet-only group lost 9.7 lbs (4.4 kg), for a difference of 2 lbs (0.9 kg). The researchers measured resting metabolic rate in two ways, and it did slow by 8-10% more in the exercise group, depending on how it was calculated (on a whole-body basis vs. per unit lean mass). The difference between groups was only statistically significant for one of these measures (per unit lean mass). Nevertheless, this difference is consistent with the 5-15% range cited in Art and Science.

Based on body composition changes, the paper estimates that over the 5-week period, the diet and exercise group lost 48,000 kcal of body energy, while the diet-only group lost 34,500 kcal of body energy, for a difference of 13,500 kcal. This is more than the amount of energy the exercise group expended in extra physical activity over the 33-day experiment (346 x 33 = 11,418 kcal). What this means is that whatever reduction may have occurred in resting metabolic rate due to exercise, it didn’t translate into a slower rate of fat loss than expected. The people who exercised lost about as much energy from their bodies (mostly as body fat) as you’d naively predict they would lose based on the number of calories they burned while exercising.

This study supports the narrow point in Art and Science that exercise slows resting metabolic rate. It doesn’t support the larger point the passage is driving toward, which is that exercise doesn’t cause as much fat loss as you’d expect based on the number of calories you burn.

Because of that, we don’t think this reference supports the passage in Art and Science very well. That said, we do want to note that the argument the book is making has received more recent support from the research of Herman Pontzer and others (see this review paper for example). It seems very possible to us that in broad strokes, the argument in Art and Science is correct – but the reduction in resting metabolic rate probably takes months or years to occur, so it wouldn’t be very visible in the cited 5-week study. It’s worth emphasizing that in this portion of the review, we’re judging how accurately the book portrays its references, rather than judging the scientific accuracy of the arguments the book is making.

Reference 5

Reference

Chapter 4, reference 26.  Jakobsen et al. Am J Clin Nutr 89:1425. 2009

Associated quote(s) and page number(s)

Page 40-41: “Moreover, three studies published in the last year have examined carefully collected dietary records of huge populations who were followed for decades. In all three of these recent studies, there was no connection between saturated fat intake and either the frequency of heart attack or death.”

Criterion 2.1. Does the reference support the claim?

2 out of 4

This reference received a score of 2 out of 4, meaning that it weakly supports the passage in the book.

The study is a meta-analysis (study of studies) that pulls together data from 11 observational studies (prospective cohort studies) on dietary intake and heart attack risk.

The results aren’t as straightforward, or as supportive, as Art and Science suggests. The book says that in this study, “there was no connection between saturated fat intake and either the frequency of heart attack or death”. In fact, there were connections in a large share of the analyses.

Rather than simply looking at the correlation between saturated fat intake and heart attack risk, this study did something called substitution analysis. This aims to estimate how risk changes when one food or nutrient is replaced by another, like replacing saturated fat with polyunsaturated fat or carbohydrate. It reports that replacing saturated fat with polyunsaturated fat is associated with a lower risk of fatal and nonfatal heart attacks. Replacing saturated fat with carbohydrate is associated with a higher risk of nonfatal but not fatal heart attacks.

The point of the passage in Art and Science is to argue that saturated fat isn’t as harmful as it’s been portrayed. These findings partially support, and partially oppose, this idea. If we believe the findings, they suggest that we’d be better off eating saturated fat than carbohydrate, but also better off eating polyunsaturated fat than saturated fat.

Reference 6

Reference

Chapter 9, reference 70.  Wang et al. Am J Clin Nutr 78:91. 2003

Associated quote(s) and page number(s)

Page 111: “Well, it has been observed that serum [palmitoleic acid] predicts the subsequent development of type-2 diabetes. In two published studies, people without any increase in blood sugar levels but increased POA are at high risk of becoming diabetic later on.”

Criterion 2.1. Does the reference support the claim?

3 out of 4

This reference received a score of 3 out of 4, meaning that it moderately supports the passage in the book.

The study measured fats in the blood in 2,909 people without diabetes (ARIC cohort) and followed them for nine years to see whether certain types of fats in the blood predict type 2 diabetes. They found that several of the fats they measured predict the risk of developing this condition. Among these, higher levels of palmitoleic acid in cholesterol esters but not in phospholipids were predictive (in models adjusted for all major confounding factors). This partially supports the passage in Art and Science, discussed further below.

We think the passage in Art and Science is a little bit inaccurate in two ways, the first more important than the second.

1. Art and Science says that the people in this study were “without any increase in blood sugar levels”, implying that high palmitoleic acid levels can predict type 2 diabetes before any sign is visible in blood sugar levels. However, the study only excluded people with blood sugar levels high enough to qualify as having diabetes, not people with mildly elevated blood sugar levels (prediabetes). In fact, table 1 reports that people who went on to develop diabetes had higher blood sugar levels at baseline, suggesting that the palmitoleic acid wasn’t necessarily providing a signal that was visible before elevated blood sugar.
2. Palmitoleic acid levels in cholesterol esters predicted type 2 diabetes, but palmitoleic acid levels in phospholipids did not. In the passage in question, Art and Science refers to “serum palmitoleic acid”, which includes both. Strictly speaking, it’s not clear that the book’s claim is supported by this study at all. However, earlier in the section Art and Science does mention cholesterol esters, without mentioning phospholipids, so a charitable reading of the text might suggest that they meant cholesterol esters. We also think this technical detail is not very important to the average reader.

Overall, we think the reference mostly supports the passage, but not fully.

Reference 7

Reference

Chapter 14, reference 114.  Kim et al. Circulation 113:1888. 2006

Associated quote(s) and page number(s)

Page 186: “Insulin resistance in endothelial cells has been shown to be due to impairment in the phosphatidyl inositol 3-kinase-dependent signaling pathway that leads to production of the potent vasodilator nitric oxide. To make matters worse, insulin also increases secretion of the vasoconstrictor endothelin-1 and this insulin signaling pathway appears to remain intact in the presence of insulin resistance.” This passage is part of the book’s argument that insulin resistance is harmful to cardiovascular health.

Criterion 2.1. Does the reference support the claim?

4 out of 4

This reference received a score of 4 out of 4, meaning that it strongly supports the passage in the book.

The reference is a review paper on the relationship between insulin resistance and endothelial dyfunction. On page 1895, it makes the aruguments described in the passage in Art and Science. This is a technical subject and we aren’t experts in it, but pathway-specific insulin resistance in general is a fairly well-established concept so we see no obvious reason to doubt the statement in the paper.

Reference 8

Reference

Chapter 13, reference 98.  Phinney et al. Arch Int Med 143:2258. 1983

Associated quote(s) and page number(s)

Page 161: “The 1970s brought us very low calorie diets (VLCD; aka protein sparing modified fasting) comprised of either common foods or prepared nutrient formulations. Most were used under medical supervision, including monitoring and provision of supplements. An exception to this practice of clinical monitoring and appropriate supplements led to the “Liquid Protein Diet” scandal – a problem of inadequate formulation and inadequate medical monitoring.”

Criterion 2.1. Does the reference support the claim?

3 out of 4

This reference received a score of 3 out of 4, meaning that it moderately supports the passage in the book.

The reference is a paper by Stephen Phinney (one of the authors of Art and Science) and colleagues, about their experience with patients on very-low-calorie weight loss diets. It reports that when 15 patients were fed high-quality protein with an adequate essential mineral intake, they didn’t experience dangerous heart arrhythmias during weight loss, regardless of the proportion of protein and carbohydrate in the diet. This is important because of the “Liquid Protein Diet scandal” the passage refers to, which the paper discusses in the introduction. In the Liquid Protein Diet scandal, 17 people died on very-low-calorie weight loss diets that used low-quality collagen protein and were very low in essential minerals, and this was thought to be related to cardiac arrhythmias caused by the diet.

The two claims related to this paper in Art and Science are that the Liquid Protein Diet scandal was caused by (1) “inadequate formulation”, and (2) “inadequate medical monitoring”. The paper directly supports (1), since it reports that a better fomulation doesn’t cause arrhythmias.

However, it’s not clear that it supports (2). The most relevant sentence in the introduction states “While careful monitoring of clinical chemistry values and metabolic balances failed to predict the arrhythmia-prone subjects in that study, appreciable depletions of whole-body potassium, magnesium, calcium, and phosphorus stores were documented for the subject group as a whole.” This quote suggests to us that better medical monitoring wouldn’t necessarily have prevented the deaths, and therefore that inadequate monitoring wasn’t necessarily a cause of the deaths. Changes in whole-body mineral stores can be measured in a research setting but it’s not realistic to take these kinds of measurements in the average patient on a medically supervised weight loss diet. The quote suggests that typical forms of medical monitoring may not have been informative.

Reference 9

Reference

Chapter 14, reference 110.  Dandona et al. Circulation 111:1448. 2005

Associated quote(s) and page number(s)

Page 185: “A large body of research has implicated elevated inflammation in metabolic syndrome and the pathogenesis of diabetes, heart disease, and other chronic diseases.”

Criterion 2.1. Does the reference support the claim?

4 out of 4

This reference received a score of 4 out of 4, meaning that it strongly supports the passage in the book.

The reference is a review paper on the connections between insulin resistance, inflammation, and metabolic syndrome, and it also discusses cardiovascular disease and diabetes. We think the contents of the paper are consistent with the passage in Art and Science.

Reference 10

Reference

Chapter 10, reference 79.  Volek et al. Metabolism 51:864. 2002

Associated quote(s) and page number(s)

Page 124: “Volek reported that normal-weight men who switched from their habitual diet (48% carbohydrate) to a ketogenic diet (12% carbohydrate) for 6 weeks significantly decreased fat mass (-3.4 kg) and increased lean body mass (1.1 kg). There was a significant decrease in serum insulin (-34%) and simple regression indicated that 70% of the variability in fat loss on the ketogenic diet was explained by the decrease in insulin concentrations.” In context, the passage appears to be arguing that the reduction in insulin that occurs on a ketogenic diet is the cause of some (maybe most?) of the reduction in body fat, rather than simply being correlated with it.

Criterion 2.1. Does the reference support the claim?

3 out of 4

This reference received a score of 3 out of 4, meaning that it moderately supports the passage in the book.

The reference is a diet trial conducted by Jeff Volek and colleagues (one of the authors of Art and Science). The details of the study are similar to what the book describes, although there are some small discrepancies. The paper describes the ketogenic diet as containing 8% carbohydrate, vs. 12% in the book. Fat loss in the ketogenic diet group was 3.3 kg, vs. 3.4 kg in the book. We don’t think these discrepancies are large enough to be misleading.

There is one thing we do think could be misleading to some readers. Art and Science states that “70% of the variability in fat loss on the ketogenic diet was explained by the decrease in insulin concentrations”, and later implies that this reduction in insulin may be a reason why low-carbohydrate diets cause fat loss. The study does report that “approximately 70% of the variability in fat loss on the carbohydrate-restricted diet was accounted for by the decrease in serum insulin concentrations”, yet it’s important to understand what that means.

What it means is that there was a strong correlation between changes in insulin levels and changes in body fat mass. When fat mass decreased a lot, insulin also decreased a lot. However, this study lacks the ability to say that the decline in insulin caused the decline in fat mass. One alternative possibility is that it’s the other way around. To our knowledge, it still isn’t known whether the decline in insulin seen with low-carbohydrate diets plays a role in causing fat loss. However, it seems unlikely that it would account for 70% of the effect, since other types of diets that don’t reduce insulin levels very much aren’t 70% less effective than low-carbohydrate diets at causing weight loss.

Overall (average) score for reference accuracy

3.3 out of 4

Summary of the health-related intervention promoted in the book

The intervention in Art and Science is dietary carbohydrate restriction. The book doesn’t provide a single specific carbohydrate target, but appears to suggest that people following a low-carbohydrate diet should eat 100 grams of carbohydrate or fewer per day (on p. 207, the book states that a hypothetical person with “pretty good carbohydrate tolerance” could eat a maintenance diet with 100 grams of carbohydrate). The sample menus on pages 231-5 are described as “providing less than 50 grams per day of total carbohydrates”.

As part of the diet, the book recommends increasing dietary fat intake, at least in the context of weight maintenance. The text and recipes don’t shy away from saturated fat.

Condition targeted by the book, if applicable

The book mostly focuses on managing metabolic syndrome, type 2 diabetes, and obesity (body fat loss). Metabolic syndrome is a cluster of features, like high blood sugar and blood lipids, that are related to insulin resistance.

Apparent target audience of the book

The back cover of the book says it’s “a great book for health-minded individuals” and “healthcare professionals”. We understand this to mean it’s partially written for average people who would like to avoid or treat obesity, metabolic syndrome, and type 2 diabetes, and partially written for healthcare providers managing those conditions. It’s a bit more technical than most general-audience nutrition books.

Criterion 3.1. Is the intervention likely to improve the target condition?

3 out of 4

We gave the Art and Science diet a score of 3 out of 4, meaning it’s likely to moderately improve the target conditions in the medium-to-long-term (6+ months).

Low-carbohydrate diets have been studied extensively for the treatment of obesity, metabolic syndrome, and type 2 diabetes, and they’re among the most effective diets for these conditions. However, in randomized trials they tend to get the best results around the 6-month mark, with benefits fading quite a bit after that (probably in large part because most people don’t stick to it very well). A non-randomized trial conducted by the book’s authors suggests that in people with type 2 diabetes, benefits can be larger and more sustained in motivated people receiving strong support.

The book’s authors also conducted a study on the impact of carbohydrate restriction on metabolic syndrome, independent of weight loss. 16 men and women with obesity and metabolic syndrome were fed three diets that were low, moderate, or high in carbohydrate, each for four weeks. Protein and calorie intake were held constant, and food was provided by the research team to help ensure the volunteers stuck to the diet.

Three features of the metabolic syndrome were related to carbohydrate intake: triglycerides, HDL-cholesterol, and blood sugar. The lower the amount of carbohydrate in the diet, the better these measurements looked. However, the intervention didn’t significantly affect other metabolic syndrome components, including blood pressure, waist circumference, and fasting insulin level (technically not part of the definition of metabolic syndrome, but thought to be related to it). That said, blood pressure and insulin trended in the right direction, and it would have been surprising to see a difference in waist circumference in four weeks on diets that all contain the same number of calories. The overall conclusion is that low-carbohydrate diets can improve certain features of the metabolic syndrome independently of calorie intake and weight loss.

Some of these benefits may be larger in settings where calorie intake isn’t controlled, because low-carb diets tend to reduce spontaneous calorie intake and cause weight loss in people with extra body fat. A meta-analysis (study of studies) on the impact of more typical low-carb diets on some metabolic syndrome components reports a larger effect of the diets on blood pressure compared with the book authors’ study, but a similar or smaller effect on HDL-cholesterol.

That’s the good news. The bad news is that these effects don’t usually last in the long run. The same meta-analysis reports that at the one-year mark, “the benefits for cardiovascular risk factors of all interventions, except the Mediterranean diet, essentially disappeared”. This problem isn’t unique to low-carb diets, but it’s nevertheless true that in the average person trying a low-carb diet, the effectiveness for weight loss and metabolic health tends to fade over time.

Criterion 3.2. Is the intervention likely to improve general health in the target audience?

3 out of 4

We gave the Art and Science diet a score of 3 out of 4, meaning it’s likely to moderately improve general health in the medium-to-long-term (6+ months).

The main conditions targeted by the book are among the most common determinants of general health in affluent countries: obesity, metabolic syndrome, and type 2 diabetes. So our reasoning is similar to the explanation for criterion 3.1.

One condition that requires special attention is cardiovascular disease. This is the most common cause of death globally, and in most countries including the US. Cardiovascular disease risk is partially, but not entirely, determined by obesity and the metabolic conditions the book focuses on.

There has been debate about the impact of low-carb diets on cardiovascular disease risk, since they tend to be high in fat, and are often high in saturated fat. This is potentially a concern due to the ability of some saturated fats to increase LDL cholesterol (“bad cholesterol”).

There is no hard evidence that low-carb diets alter the risk of cardiovascular disease in either direction, because there are no low-carb diet trials with cardiovascular outcomes (at least, none with enough events to be informative). That said, in general, we aren’t very worried about the impact of low-carb diets on LDL cholesterol. First, low-carb diet trials mostly don’t report that the diet increases LDL cholesterol in the first place, at least in people with obesity, overweight, and/or type 2 diabetes who are losing weight. Trials of the low-carb Atkins diet do report an increase in LDL-cholesterol, but it tends to be small. Second, LDL cholesterol isn’t the only determinant of cardiovascular risk, and other cardiovascular risk factors like blood pressure tend to improve on low-carb diets.

That said, some low-carb diet trials, and some case studies, do report large increases in LDL cholesterol. Even though this doesn’t appear to be typical, it may be wise to measure blood lipids after transitioning onto a low-carb diet, and talk with a doctor if anything looks concerning.

Speaking of cardiovascular risk, Art and Science advises low-carb dieters who eat fewer than 60 grams of carbohydrate per day to “purposefully add 2-3 grams of sodium to your daily intake” (p. 241). The reason for this is that “removing most carbs from the diet causes the kidneys to aggressively secrete sodium (and along with it, excess fluid)” (p. 240). Although the book makes this point many times, and it’s a common claim in low-carb circles, the book doesn’t cite references to support it.

While we don’t doubt the authors’ clinical experience that patients feel better when they increase their salt intake, a higher intake of salt is potentially concerning due to its link with high blood pressure and cardiovascular events like heart attacks and stroke.

We wondered whether there is evidence to support the claim that low-carbohydrate diets increase sodium excretion. A quick search identified three low-carb diet trials with fewer than 60 grams per day of carbohydrate that report sodium balance or excretion. Of these, only one reports that low-carb dieters lost more sodium than higher-carb dieters, and that effect was only present in the first week of the 28-day diet. The paper, published in 1981, states that “diets restricted in carbohydrate cause an initial increase in sodium excretion”, citing previous studies (emphasis ours).

It appears that low-carbohydrate diets may increase sodium excretion initially, but the body compensates for it fairly quickly. Yet Art and Science doesn’t mention that people should stop adding extra salt to the diet after the first week.

Most people probably eat too much salt. The evidence on this remains somewhat uncertain because we don’t have large salt reduction trials designed to detect changes in cardiovascular events, but reducing salt intake does tend to reduce blood pressure and salt intake is correlated with cardiovascular risk in the highest quality observational studies available.

In the US, typical intake is 3.4 grams of sodium per day, and the recommended intake is less than 2.3 grams per day. Art and Science recommends low-carb dieters increase their sodium intake by 2-3 grams per day. If readers begin from an average intake, that would mean a total of 5.4-6.4 grams per day for the average American eating the diet. This is 2-3 times the recommended intake of salt.

It’s theoretically possible that a high-salt diet isn’t harmful to people eating a low-carbohydrate diet, but we haven’t seen evidence of this. We are somewhat reassured that people with obesity or overweight who begin a low-carb diet usually experience reduced blood pressure. This might counteract the possible harm of the high intake of salt. That said, two things leave us with some concern. First, the harms of salt may not all be via its impact on blood pressure. Second, most people usually regain most of the weight they lose on a low-carb diet, and some of the damage caused by a high intake of salt may persist even after salt intake declines.

Finally, we think it’s important that the diet would cause people to eat less of the types of food we think are the most problematic in typical diets. These are the highly processed, calorie-dense, seductive items that go by various names like ultra-processed food and junk food. We think these are an important driver of obesity, metabolic disease, and cardiovascular disease, and limiting them is likely to improve health in most people.

Criterion 3.3. Does the diet portion of the intervention promote an adequate nutrient intake for general health in the target audience?

3 out of 4

We gave the Art and Science diet a score of 3 out of 4, meaning we think it’s more than nutritionally adequate. In other words, we think it provides enough of all essential nutrients (except vitamin D, which can be obtained from sunlight or supplements), and somewhat more health-promoting non-essential nutrients than typical diets.

To score the diet’s nutrient adequacy, we first looked for data on low-carb diets in general. We found a study comparing the self-reported nutrient intake of 307 US adults on a low-carb diet vs. people not on a diet. The results suggest that low-carb dieters typically eat inadequate amounts of several vitamins, minerals, and dietary fiber. Most of the nutritional shortcomings are related to insufficient intake of whole plant foods: fiber, potassium, magnesium, and vitamin C. However, low-carb dieters had slightly better nutrient intake than people not on a diet.

That said, we think people who follow the guidance in Art and Science should do better than this. Pages 232-235 provide a 7-day meal plan and it includes ample whole plant foods. For example, day 3 includes servings of cauliflower, mixed nuts, lettuce, sorrel, cucumber, and tomato. We haven’t run the meal plan through a nutrient estimator, but we think the diet as presented in the book is more than nutritionally adequate.

That said, we think this meal plan is ambitious for the average person, given that it involves a lot of cooking, even for breakfast. Will most readers apply it faithfully, or will their motivation wane over time, in favor of easier and less nutritious low-carb foods? We’d guess somewhere in the middle for most people.

Overall (average) score for healthfulness

3 out of 4