Most research indicates little difference in body composition changes between carbohydrate or fat restriction between isocaloric, isonitrogenous diet groups. Those with impaired glucose metabolism/insulin resistance may be more susceptible to weight regain on higher-carbohydrate diets, but this population is unique. In general, low-carbohydrate/ketogenic diets produce similar fat loss when compared to isocaloric, high-carbohydrate diets.
Significant research examining this question
Metabolic Ward Trial 1:
The most tightly-controlled (2015) crossover trial comparing the two diet conditions (68) was conducted with 19 obese individuals in a metabolic ward (a laboratory where participants stayed) over 4 weeks. This setting allowed researchers to control energy, carbohydrate, protein, fat, fiber, saturated fat, and sugar intakes for all subjects in both groups. Participants were randomized and fed a diet matched to their respective daily energy expenditures (as measured in the laboratory). Subjects then spent 6 days on either a low-carbohydrate or high-carbohydrate, higher-sugar diet with calories and protein matched.
Researchers took measurements and repeated the process 2-4 weeks later, swapping each group’s diet the second time. Body composition (via DXA) and insulin secretion were measured in both diet groups over time. The authors outline, “This study demonstrated that, calorie for calorie, restriction of dietary fat led to greater body fat loss than restriction of dietary carbohydrate in adults with obesity. This occurred despite the fact that only the carbohydrate-restricted diet led to decreased insulin secretion…”
Therefore, under strictly controlled laboratory conditions with energy intake matched, a higher-carbohydrate (thus, insulinogenic) diet produced slightly greater fat loss than a low-carbohydrate diet. The researchers continue, “…we can definitively reject the claim that carbohydrate restriction is required for body fat loss (Taubes, 2011).” All things considered, energy intake seemingly predicts fat loss outcomes better than carbohydrate intake does, despite insulin’s lipogenic function.
Some criticized this study, claiming that, because of its highly rigorous control, it failed to capture the fat loss advantage that would have prevailed with greater carbohydrate restriction in less controlled settings (given practical benefits of lower carbohydrate diets). Others argued that the study’s results must be replicated, given the study’s small sample size and short duration. Furthermore, people argued that the study’s results aren’t practically useful, because participants weren’t fed “healthy” low-carbohydrate diets.
The DIETFITS Trial:
Fortunately, the (2018) DIETFITS trial (69), bearing a much larger sample (n=609), longer duration (12 months), “healthier” food selections, and less control (non-metabolic ward), confirmed the former study’s findings. This study compared high and low-carbohydrate diets (supervised by health educators) to determine if genotype or insulin secretion influence weight loss.
While neither baseline insulin secretion nor genotype affected outcomes, the researchers also found no significant differences in body weight, waist circumference, or body fat changes between groups. The latter three results further support the finding that fat loss is similar when protein and calorie intakes are controlled (63,66,88,92–100,185). Again, different insulin levels produced similar fat loss. This renders Taubes’ conclusion questionable, as carbohydrate-induced insulin secretion doesn’t impede short-term (1 week-12 months) fat loss (68,69,185).
Metabolic Ward Trial 2:
Additionally, researchers conducted another metabolic ward trial in 2016 (186), comparing a ketogenic diet (<50 grams of carbohydrates per day) to an isocaloric, higher-carbohydrate diet. This 2-month study of 17 overweight (or obese) males again confirmed that fat loss did not differ between diets.
The best available research demonstrates similar fat loss between fat and carbohydrate-restricted diet groups. Both the most tightly-controlled research and a well-designed longer-term, large-sampled trial reported similar fat losses between diet conditions. This indicates that greater carbohydrate restriction fails to yield greater fat loss.
Does sugar produce detrimental health outcomes/fat gain?
Even if Taubes’ hypothesis is false, sugar may still be fattening/unhealthy. If Taubes’ concerns are valid, then research should demonstrate sugar’s ill effects.
Sugar and health
A 2018 review (120) investigating sugar’s metabolic effects, among other points, highlighted the following:
- Our bodies need glucose for many of its functions (e.g. neuronal function).
- Dietary sugar consumption up to 80% of one’s energy intake seems safe in physically active populations (e.g. the Hadza (79–81), Kuna (82), Mbuti (122), and Ache (123)).
This suggests most sugar intakes (within one’s calorie needs) are likely safe, as mentioned populations rarely experience metabolic health issues.
Sugar In Vegetables and Fruits:
Moreover, carbohydrates/sugars are prevalent in vegetables and fruits, which consistently correlate with improve health markers (144–146,229) (perhaps due to their (often) high antioxidant (148,152), vitamin (149), mineral (149), phytochemical (150,152), and polyphenol (151) contents). Were Taubes correct that sugar is “…the principal cause of the chronic diseases that are most likely to kill us, or at least accelerate our demise…”, then sugar-rich vegetables and fruits would be harmful, or their physiological effects would have to somehow counteract insulin secretion. Contrasting Taubes’ claim further, sugar-rich mangoes don’t seem to induce weight gain (268) and provide many health benefits (147), despite their insulinogenic effect. The above seemingly contradicts Taubes’ claim that sugar is responsible for obesity and chronic diseases.
Fructose and Health:
The scientific literature currently debates fructose’s (a type of sugar) health impacts. Some researchers suggest the safe upper limit to fructose intake may be as low as 25-40 grams per day (89), warning that excess fructose consumption may increase apolipoprotein-B concentrations (250) and raise blood pressure (at least acutely) (249). This seemingly supports Taubes’ recommendation to limit sugar intake.
Contrastingly, others suggest up to 100 grams of fructose per day is safe in humans (90) (when examining effects on plasma triglycerides, weight, and HbA1c). Additionally, a recent review (117) assessing fructose vs. glucose, and sucrose vs. high-fructose corn syrup’s inflammatory effects reported no significant differences between sugars. Each sugar similarly affected inflammation, with no significant effect on inflammation in any case (based on limited evidence, however).
Further, another meta-analysis (119) reported that swapping fructose for other carbohydrates doesn’t contribute to fatty liver disease, unless it coincides with excess calories (a confounder, as excess energy intake causes fat gain) (187,201). A 2008 survey (121) estimated mean fructose consumption to be 54.7 grams per day (range: 38.4–72.8) among Americans, with adolescents (12–18 years) consuming 72.8 grams per day. This intake appears relatively safe (141,202), as a 2010 review (202) reports, “…moderate fructose consumption of ≤50g/day…has no deleterious effect on lipid and glucose control and of ≤100g/day does not influence body weight.”
Lastly, a 2017 review (252) concludes that “Although it seems apparent that increased intake of fructose leads to various risk factors associated with metabolic syndrome…there is still numerous contradictory evidence which states that as long as fructose is consumed in moderate doses, fructose may not augment these risk factors.” The above evidence suggests that moderate fructose consumption (perhaps ≤80 grams per day) seems safe.
Even if Taubes’ hypothesis is false, sugar may still be fattening/unhealthy. If Taubes’ concern that sugar causes diabetes and obesity is valid, then research should demonstrate sugar’s ill effects. A 2018 review highlights that our bodies actually need sugar, and that many populations safely consume up to 80% of their energy intake from sugar. Moreover, carbohydrates/sugars are prevalent in vegetables and fruits, which are consistently shown to improve health markers. This contradicts Taubes’ claim that sugar intake conduces to disease. While researchers debate fructose’s health impacts, studies suggest that moderate fructose consumption (≤80 grams per day) is probably safe.
Sugar and fat gain
Even if carbohydrate fails to induce fat gain, sugar may still cause fat gain (despite not degrading health). Taubes writes, “The purpose of this book is to present the case against sugar–both sucrose and high-fructose corn syrup…” Seemingly supporting this case, a 2018 review (91) asserts, “Added sugar consumption in early life is associated with higher obesity in childhood.” This association between sugar consumption and obesity suggests Taubes may have a point.
While this association may generally hold true, it is true of added sugar specifically, rather than sugar indiscriminately. There’s a clear distinction here, as sugar-containing fruits/vegetables yield better health outcomes than more processed, less nutritious foods with added sugars (171). I don’t recall Taubes making such a distinction. Nonetheless, this paper merely reports an association, which doesn’t imply causation (especially since energy intake confounds sugar intake). Contrastingly, (sugar-containing) fruit consumption often inversely correlates with obesity (264–266).
Even if sugar generally correlates with obesity, this isn’t always true, as the opposite association was documented in China (83–87). Additionally, the review’s authors actually cede that sugar’s effects are potentially obesogenic given their influence on energy balance in writing: “The high-sugar…diet could be perturbing both sides of the energy balance equation…” (91). Since these authors analyze sugar’s effects in the context of energy balance, they seem to disagree with Taubes’ suggestions.
Further, the aforementioned “significant” research (68,69,186) (see Significant Research Examining This Question), maintained higher sugar intakes in the high-carbohydrate diet groups. Since research finds no major fat loss differences between high and low-sugar diets, sugar’s correlation with obesity is conditional. It’s thus unlikely that high added sugar intake causes childhood obesity. Rather, high sugar intake is likely just one of many factors potentiating high energy intake, which drives obesity.
Fructose and Fat Gain:
One detriment to fructose consumption is its increased potential for liver/visceral fat storage (124,177,178), and ability to decrease insulin sensitivity as compared to glucose (124). Yet, these outcomes only occurred given a caloric surplus. Therefore, positive energy balance confounds any of sugar consumption’s (mentioned) ill effects (187,201).
Moreover, fructose doesn’t uniquely favor liver/visceral fat storage during overfeeding. Of note, saturated fat also demonstrates increased potential for liver/visceral fat storage (175,176), despite not affecting insulin secretion (still, only given positive energy balance). This point merely highlights that even if fructose (thus, sugar) overconsumption is harmful, it’s no more detrimental than saturated fat overconsumption (in this regard). Interestingly, saturated fat overfeeding may increase liver fat more than overfeeding on sugars (251).
Even if carbohydrate fails to induce fat gain, sugar intake may still cause fat gain (despite not degrading health). A 2018 review reports early-life added sugar consumption correlates with higher childhood obesity. Yet, this evidence is weak because correlation is not causation, added sugar isn’t naturally-occurring sugar, and energy intake confounds sugar intake.
Even if sugar generally associates with childhood obesity, this isn’t always true, as the opposite association exists in China. Positive energy balance confounds many of sugar/fructose consumption’s detrimental effects. For example, (sugar-containing) fruit consumption often inversely correlates with obesity. While fructose overfeeding stimulates more liver/visceral fat storage than glucose overfeeding, it might cause less liver/visceral fat storage than saturated fat overfeeding.
Is there evidence against the notion that sugar causes obesity?
While body composition changes similarly when researchers match protein and energy intakes, this doesn’t directly counter sugar’s obesogenic potential. Despite the DIETFITS trial (69), some still question the degree to which carbohydrates might hinder fat loss/cause fat gain in “real-world” settings, over longer durations. If Taubes’ hypothesis is correct, then we should neither find healthy populations with high sugar intakes, nor obese populations with low/decreasing sugar intakes over multiple years.
Healthy populations with high sugar intake
There are observational data from several cultures contradicting Taubes’ hypothesis that higher carbohydrate/sugar diets are inherently more fattening. Note the foraging Hadza of Tanzania. They often consume honey, berries, and tubers (all high-carbohydrate foods) (79). Honey is their preferred food source (80,81), while they primarily consume berries (79), and tubers are their least favorite food (80,81). The Hadza, despite eating many calories from carbohydrate/sugar, maintain low average body fat percentages (~11% for males, ~20% among females) (81) (and rarely experience obesity (184)). The Kuna Indians of Panama also remain healthy despite consuming many carbohydrates from fruit (82).
Obese populations with decreasing carbohydrate and sugar intake
Additionally, carbohydrate’s absolute (and relative) contribution to the Chinese diet decreased from 1991 to 2011 (from 387-404 grams/day to 315-328 grams/day) (83–85), while obesity simultaneously increased (83–87). Sugar intake also generally declined over the same time period (263). Moreover, energy intake predicted Chinese obesity tends better than carbohydrate/sugar intake. Similarly, Chinese obesity was positively related to fat/protein consumption, while negatively associated with carbohydrate/sugar intake. (Edit: see “November 2018 edit” section). These data counter the notion that carbohydrates/sugars cause obesity, as carbohydrate inversely correlates with obesity in China, and sugar hasn’t caused obesity in the Hadza/Kuna (82)/Mbuti (122)/Ache (123) population.
If Taubes’ hypothesis is correct, then we shouldn’t find healthy populations with high sugar intakes, nor obese populations with low sugar intakes over long durations. The Tanzanian Hadza/Indian Kuna are foraging populations with high sugar intakes, yet low obesity rates. Additionally, despite China’s decreasing carbohydrate (and sugar) intake between 1991 and 2011, obesity rates increased, thus correlated negatively with carbohydrate/sugar intake. These two cases seemingly contradict Taubes’ hypothesis.
Consider alternative hypotheses
The alternative hypothesis to Taubes’ claim is that carbohydrates are not inherently more fattening than other (isocaloric) nutrient sources. The “energy balance” concept (also known as calories in; calories out/CICO) is the prevailing paradigm. The first thermodynamic law certifies that energy (calories) can neither be created nor destroyed; only transformed from one form to another (37). From a thermodynamic perspective, a calorie is a calorie (48), irrespective of its macronutrient source.
State the claim
Since humans are thermodynamic systems (38,39), changes in body energy stores depend on an energy imbalance (40–42). Therefore, if you consume more calories than you burn, you increase the body’s energy stores (and vice versa). Changes in bodily energy stores coincide with changes in body mass (i.e. water, protein, glycogen, and fat mass (43–46)). Considering the above, it follows that changes in bodily energy stores depend on energy intake, rather than carbohydrate/sugar intake.
There is scientific consensus that energy balance dictates body weight change (49,185). For example, here are two study collections (400+ total studies) supporting the notion that changes in energy balance determine weight loss/gain (50,51). However, while energy balance primarily determines weight change, protein intake alters this weight’s composition.
Foremostly, protein intake tends to decrease fat gain, improve fat loss, or increase lean mass gain (66,71,101,102,118). Thus, protein alters the percentage of body fat loss/gain (unlike carbohydrate/fat intake) when one consumes ~1.6-3.4 grams of protein per kilogram (~.73-1.54 grams per pound) of bodyweight (108,71). Lean mass gains are more probable if said protein intake is paired with resistance exercise (102,106–111,118).
Further, protein intake uniquely shifts the energy balance equation, given its greater thermogenic effect (66,71,101,104,105,142,162) (its metabolism burns more calories), reduced energy efficiency (66,71,104) (it’s more likely to be oxidized or otherwise excreted, rather than stored), hunger-blunting effect (112,113), satiety-inducing effect (101,103,104), and consequent ability to increase energy expenditure while reducing energy consumption (66,114–116). Considering the above, protein intake improves body composition unlike other macronutrients, given its physiological effects. Therefore, research must match protein intake when comparing different carbohydrate and fat intakes, for its consumption considerably influences body composition (among other confounders).
The alternative hypothesis states
Claim: When energy balance (calories) and protein intake are held constant, carbohydrate/sugar consumption is as fattening as fat consumption, regardless of carbohydrate/sugar’s physiological effects (even when consumed chronically, over long durations). However, protein affects (improves) muscle:fat mass ratio (body composition) unlike other macronutrients, given its physiological functions and effects on energy balance.
The hypothesis proclaims
Premise 1: The first thermodynamic law ensures that when energy passes into or out of a system, the system’s internal energy changes in accordance with the law of conservation of energy.
Premise 2: Changes in body energy stores depend on energy imbalance. Therefore, if you consume more calories than you expend, you increase the body’s energy stores and vice versa.
Premise 3: An increase in the body’s energy stores over time causes body mass gain (in the form of fat mass or lean body mass (body weight-body fat mass)).
Premise 4: Protein sources are highly satiating, while yielding an increased thermogenic effect, decreased energy efficiency, and greater muscle mass (for equivalent caloric intake) as compared to carbohydrate or fat. Additionally, consuming more protein when overeating may produce less fat gain, despite similar muscle growth. As such, protein consumption improves body composition (lean mass:fat mass) unlike other nutrients, given its physiological effects and influence on energy balance.
Conclusion: A (massive) positive energy balance (not carbohydrate/sugar intake) over time causes obesity, for this energy surplus is mostly stored as fat mass, though protein intake can improve the ratio of lean:fat mass gain.
Examine the evidence for the claim
Considering all aforementioned evidence, this hypothesis is plausible. See section titled “Does greater carbohydrate restriction yield more fat loss?” for a summary of the strongest corroborating evidence.
Rate, according to the criteria of adequacy, each hypothesis
Testability: Is the hypothesis testable?
Both hypotheses are testable, yet Taubes’ exact hypothesis predicts that sugar’s effects take years to culminate in obesity (just as smoking often takes years to cause cancer). Meanwhile, researchers can successfully test the energy balance hypothesis over any time period (thus, more easily). Indeed, Taubes grants, “Thousands if not tens of thousands of subjects have to be randomized to high- and low-sugar diets and then followed for years (the more subjects in the study, the shorter the trial needs to run) to see which group experiences the greater toll in sickness and death…” in order to truly test his hypothesis. The energy balance hypothesis is thus more testable.
Advantage: Energy Balance
Fruitfulness: Does the hypothesis yield observable, surprising predictions that explain new phenomena?
Taubes’ hypothesis does not seem to explain obesity any better than the energy balance hypothesis. Taubes’ hypothesis is less fruitful, because it cannot explain why insulinogenic protein consumption improves body composition. It also cannot explain childhood obesity’s prevalence, if he insists that sugar’s ill effects take years to manifest.
Further, Taubes’ hypothesis cannot explain the similar fat losses observed in the DIETFITS or metabolic ward studies (68,69,186) (with differing carbohydrate/sugar intakes and insulin levels between groups). Likewise, the insulin hypothesis neither explains high-sugar fruits (e.g. mangoes (147)) and vegetables’ health benefits (144–146) nor fruit’s anti-obesogenic effects (264–266). Meanwhile, these benefits don’t contradict energy balance. Lastly, Taubes’ hypothesis cannot explain the low obesity rates among the sugar-loving Hadza (nor other mentioned populations).
The energy balance hypothesis yields more observable, surprising predictions, as it can explain all of the above. Moreover, the energy balance hypothesis explains why insulin cannot cause fat gain without an energy surplus (136), despite its lipogenic function.
Advantage: Energy Balance
Scope: How many different phenomena can the hypothesis explain?
The energy balance hypothesis has much greater scope, as the physical laws of thermodynamics and conservation of energy apply universally. They explain heat transfers (e.g. conduction, convection, and radiation) (143), climate change (154), and a perpetual-motion machine’s impossibility (1). Conversely, Taubes’ hypothesis only explains human fat storage, however inadequately (as mentioned above).
Advantage: Energy Balance
Simplicity: Is the hypothesis the simplest explanation for the phenomenon?
The energy balance hypothesis is simpler. It only assumes that humans are subject to physical laws (and that physical laws exist). Taubes’ hypothesis assumes that sugar consumption/insulin secretion defies physical law, and that sugar’s ill effects take years to manifest (like tobacco). The calories in; calories out model assumes less, and is therefore, simpler.
Advantage: Energy Balance
Conservatism: Is the hypothesis consistent with our well-founded beliefs/empirical evidence?
There is scant evidence supporting Taubes’ hypothesis, while there is abundant favoring energy balance (50,51,63,66,78,88). Thus, the latter hypothesis is more consistent with empirical evidence. The energy balance model is also more consistent with our well-founded beliefs (153), because it agrees with physical laws, vegetables/fruits’ health benefits (144–146,148–152,264–266), carbohydrate/sugar consumption’s negative association with Chinese obesity (83–87), the Hadza’s low obesity rates (81), and bodily sugar needs (120). All told, the energy balance hypothesis is more conservative.
Note: Energy balance is more of a scientific theory, since its application spans many disciplines (153,154).
Advantage: Energy Balance
In conclusion, little evidence supports Taubes’ claim that sugar consumption causes obesity. Simultaneously, a plethora of research favors the energy balance hypothesis. The thermodynamic model also scores better in the criteria of adequacy; superior in conservatism, scope, fruitfulness, simplicity, and testability. As such, it is reasonable to reject Taubes’ hypothesis in favor of the energy balance model.
All things considered, cutting your sugar intake alone may fail to elicit fat loss. If you want to ensure fat loss/muscle gain, you can learn all about diet setup, calorie tracking, and body recomposition here.
November 2018 edit
Sugar intake may not have decreased in China during obesity’s rise after all. While China’s total carbohydrate intake decreased over this time period, and the rural data in the previously cited article (263) suggested that sugar intake may have decreased, we cannot say for sure. This is due to migration from rural to urban China and population growth. Notice that the Chinese population increased by tens of millions over this time period (269), the rural population decreased by hundreds of millions, and China’s population demographic dramatically shifted from mostly rural to mostly urban (270).
The above indicates that, while the sugar consumption data were being collected for the rural population, this population was actually decreasing by tens of millions, while the urban population was increasing even more dramatically. This means that the rural Chinese sugar consumption data bore diminishing contribution to China’s total sugar intake past the year 2006, and as such provides little insight into the true Chinese sugar consumption levels.
Furthermore, the referenced article showing China’s urban and rural sugar consumption (263) is limited, as it only assessed at-home sugar consumption among households, rather than total sugar consumption across the households and factories. These data are thus especially limited.
Lastly, other research reports increased sugar harvests over the time period that the referenced paper shows sugar consumption decreasing (271). Since sugar production and harvesting increased, it implies that sugar use/consumption increased in China. This doesn’t necessarily contradict the other paper referenced, as the other paper could not account for away-from-home sugar consumption.
Nonetheless, added sugar’s correlation with U.S. obesity may have still been negative from 2003-2012. While absolute (added) sugar intake decreased from 2003 to 2012 (207), obesity continued to increase (204). I say the correlation may have been negative because the referenced paper (207) relies entirely upon US survey data. Granted, the data are from 6 large cohorts, but self-reported food intake is consistently unreliable.
For example, one review (272) suggests that memory-based dietary assessments “bear little relation to actual energy or nutrient consumption”, while others note that people usually under-report their food intake (273, 274)
Frequently asked questions
Is sugar addictive?
While evidence suggests that sugar isn’t addictive (192–195), food may be an addictive substance (192,195–197), given its potential to stimulate brain-reward dysfunction (192,197–199).
There is some debate as to whether food is an addictive substance, or eating is an addictive behavior (195,197). A 2018 meta-analysis (197) concludes, “Though both behavioral and substance-related factors are implicated in the addictive process, symptoms appear to better fit criteria for substance use disorder than behavioral addiction.” Whether or not food/eating is physiologically addictive, human eating behaviors might resemble addiction.
However, sugar isn’t likely addictive (193,194), and no macronutrient seems addictive on its own (195). This is plausible, as people don’t typically binge on granulated sugar/liquid oil/amino acid powder in isolation. Yet, they might over-consume chocolate, cookies, or pizza (some combination of the macronutrients).
Rather than sugar content, energy density seemingly dictates desire for certain foods (192,199). Additionally, individual eating preferences contribute to any specific food’s reward value (192,198), and food preferences might be conditioned through experience (198). This information is consistent with Hadza behavior, as their favorite food (on average) is honey (80,81), which is both the most calorically-dense option and a food they’ve learned to enjoy.
More specifically, evidence suggests addictive foods are usually highly-processed, high-glycemic, rife with added fats and/or refined sugar (sweet/savory), and highly palatable (192,196,197). However, while this generally holds true, there are exceptions to this guideline. For example, nuts (whole food without added sugars) have been rated to be more addictive than granola bars (processed food with added fat and sugar) (196). Though this is anomalous, even carrots (unprocessed vegetables) have documented addictive properties (200).
One paper (196) identified the following foods among the most potentially addictive: Pizza, chocolate, ice cream, chips, cookies, french fries, cheeseburgers, bacon, cake, popcorn, muffins, soda, cheese, cereal, fried chicken, rolls, and gummy candy. If you exhibit addiction-like behaviors for any of these foods, or tend to over-consume them, then consider limiting consumption. Individual preferences will vary, but in avoiding “food addiction”, limiting your exposure to problem foods should help.
While evidence suggests that sugar isn’t addictive, food may be an addictive substance, given its potential to stimulate brain-reward dysfunction. There is some debate as to whether food is an addictive substance, or eating is an addictive behavior. A 2018 meta-analysis concludes, “Though both behavioral and substance-related factors are implicated in the addictive process, symptoms appear to better fit criteria for substance use disorder than behavioral addiction.” However, sugar isn’t likely addictive, and no macronutrient seems addictive in isolation.
Rather than sugar content, energy density seems to dictate desire for certain foods. More specifically, evidence suggests that addictive foods are often highly-processed, high-glycemic, rife with fats and/or sugar (sweet/savory), and highly palatable. However, while this generally holds true, there are exceptions to this guideline.
Why might sugar consumption seem so harmful?
It might seem like obesity and carbohydrate/sugar intake associate more closely in China, however what seems doesn’t always match reality (1). Human memory is subject to many biases and fallacious recollections, as memory is both imperfect and reconstructive (1). While our anecdote indicates that Chinese carbohydrate intake is increasing, our experience can easily misjudge a phenomenon’s prevalence.
For example, we might remember times when the obese consumed sugary foods. This makes it seem like sugar causes fat gain (false cause). Unfortunately, our memory often fails to capture the “full picture”.
While we’re likely to remember instances in which sugar consumption seemingly produced fat gain, we might ignore counterexamples. We may recall an instance of (sugary) ice cream-induced fat gain, yet simultaneously neglect (low-sugar) bacon-induced fat gain. Similarly, how can we know that sugar is the culprit, when ice cream is also high-fat and calorically-dense?
Indeed, we’re susceptible to cherry picking evidence (consciously or not), confirmation bias (searching for, interpreting, favoring, and recalling information to support pre-existing beliefs), availability error (recalling more salient outcomes over others), reliance on anecdotal evidence (problematic when data exist), overlooked causes/exclusion (neglecting crucial information in making an argument), denial (rejecting contrary evidence), and investigator bias (seeing a desired/expected effect, despite its absence), among many other fallacious tendencies (1).
Considering the above, it’s easy to believe that sugar causes obesity. We might rely upon anecdotal/observational evidence suggesting carbohydrate’s/sugar’s link to obesity (206), while overlooking other relevant scientific data (anecdotal evidence, then exclusion or denial). Additionally, we could recall one obese person’s addiction to (sugary) soda, then continue seeking corroborating cases (cherry picking, availability error, or recall bias, then confirmation bias). Further, we could plausibly overlook positive energy balance as an alternative cause for obesity. All the while, we may neglect obese people over-consuming oils (fat), as our expectation of sugar’s guilt blinds us (recall bias, availability error, exclusion, or denial, then investigator bias).
It’s prudent to beware of memory’s tremendous limitation and our susceptibility to biases, as these can fatally confound our judgments. In this article, I attempted to seek contrary evidence and avoid strawmanning. Nonetheless, I may have cherry-picked citations and unintentionally weakened Taubes’ argument. The reader must take care to analyze all relevant evidence in reaching conclusions.
What seems doesn’t always match reality. Human memory is subject to many biases and fallacious recollections, as memory is both imperfect and reconstructive. Indeed, we’re susceptible to cherry picking evidence (consciously or not), confirmation bias (searching for, interpreting, favoring, and recalling information to support pre-existing beliefs), availability error (recalling more salient outcomes over others), reliance on anecdotal evidence (problematic when data exist), overlooked causes/exclusion (neglecting crucial information in making an argument), denial (rejecting contrary evidence), and investigator bias (seeing a desired/expected effect, despite its absence), among other fallacious tendencies. It’s prudent to beware of memory’s tremendous limitation and our susceptibility to biases, as these can fatally confound our judgments. In this article, I attempted to seek contrary evidence and avoid strawmanning. Nonetheless, I may have cherry-picked citations and unintentionally weakened Taubes’ argument.
Addressing sugar’s impact on obesity in the United States
Obesity is undoubtedly widespread in the United States (204–206). Additionally, carbohydrate’s relative dietary contribution has increased both generally and among the overweight (206). While much of this carbohydrate seems to come from (added) sugar (207), correlation is not causation, and there are other noteworthy factors.
Foremostly, while carbohydrate intake and obesity correlate positively, carbohydrate’s relative dietary contribution increased by merely 4.7% between 1971 and 2006 (206), while obesity rates simultaneously increased by 21.5% in men and 19.9% in women (206). Carbohydrate intakes among overweight men and women only increased by ~56 and ~53 grams (on average) respectively, over this time period (206). Obesity wouldn’t likely have increased so dramatically if its sole cause was carbohydrate intake, as carbohydrate intake didn’t simultaneously increase substantially.
Ironically, added sugar’s correlation with U.S. obesity is actually negative from 2003-2012. While absolute (added) sugar intake decreased from 2003 to 2012 (207), obesity continued to increase (204). Therefore, neither carbohydrate nor sugar intake seem to cause obesity alone.
Unsurprisingly, energy intake better correlates with obesity (204,207). However, while positive energy balance causes obesity mechanistically, other factors may promote obesity in practice. For example, processed food consumption is associated with excess weight and poor diet quality (208–210). Processed food consumption’s link to excess weight is plausible, for (as mentioned above) processed foods are addictive, energy dense, and hyper-palatable, thus conducive to overconsumption (209). Further, processed food’s contribution to obesity may fuel the misconception that carbohydrate/sugar consumption causes obesity, as 89% of added dietary sugar comes from processed food (208).
Unfortunately, hyperpalatable processed foods (often high in fat and sugar) are very cheap (209), while greatly rewarding (acutely), which promotes chronic (over)consumption. In addition to being hyperpalatable, energy dense, fat/sugar-rich, and addictive, processed foods seem inherently more fattening (211).
Indeed, as compared to an isocaloric, less-processed meal, a more-processed meal decreased postprandial energy expenditure by 50%, indicating its potential to increase obesity risk by diminishing energy output (211). All things considered, reducing processed food consumption is likely a good way to improve the U.S.’ diet quality (210), as processed foods decrease energy expenditure (211), lack micronutrients (210), and promote increased energy intake (see above), favoring weight gain over time.
Obesity is undoubtedly widespread in the United States. Additionally, carbohydrate’s relative dietary contribution has increased both generally and among the overweight. Ironically, added sugar’s correlation with U.S. obesity is actually negative from 2003-2012. While absolute (added) sugar intake decreased from 2003 to 2012, obesity continued increasing. Therefore, neither carbohydrate nor sugar intake seem to cause obesity alone. Unsurprisingly, energy intake better correlates with obesity. However, while positive energy balance causes obesity mechanistically, other factors may promote obesity in practice. Food availability, processing, addictive potential, and other factors contribute to the positive energy balance pushing obesity forth.
Are there any good reasons to avoid carbohydrates/sugar?
While carbohydrate/sugar is neither inherently more fattening, addictive, nor unhealthy (given reasonable consumption), this doesn’t imply absence of reason to avoid consuming sugar. In fact, you may benefit tremendously from limiting your sugar intake, but potential for benefit is highly individual.
One 2018 review (91) outlines, “Emerging evidence suggests that, following weight loss on energy-restricted diets, ad libitum consumption of a high-CHO/high-glycaemic-load diet may, via increased insulin exposure, decrease insulin sensitivity in muscle and increase insulin sensitivity in adipose, thus increasing susceptibility to weight regain.” Thus, carbohydrate restriction may benefit the insulin resistant by lowering their susceptibility to weight regain.
Interestingly, multiple studies report greater fat loss among the less insulin sensitive on low-glycemic load diets (212–215), as well as improved adherence (216), increased energy expenditure (275), and health markers (217) on low-carbohydrate diets (though, not always (68,69,186,218,219)). Supporting these findings, some research suggests that insulin sensitive individuals lose more fat on higher carbohydrate diets (220,221) (but again, not always (69,218,219)).
The inconsistent findings suggest that individuals respond dynamically, as supported by some research indicating genetic influence on response to high/low-carbohydrate diets (222–225). However, these findings are also questionable (69), and more research is needed. Though effects seem individual, the general trend indicates greater fat loss with greater carbohydrate restriction/lower glycemic load among insulin resistant dieters (91). The insulin resistant should thus (generally) avoid higher glycemic load or carbohydrate-dense food sources (see glycemic loads of various foods here (226)).
One 2018 review outlines that carbohydrate restriction may benefit the insulin resistant by lowering their susceptibility to weight regain. Interestingly, multiple studies report greater fat loss among the less insulin sensitive on low-glycemic load diets. Similarly, papers report improved adherence and health markers when the insulin insensitive try low-carbohydrate diets (though, not always). The inconsistent findings suggest that individuals respond dynamically, while research indicates genetic influence on response to high/low-carbohydrate diets. However, these findings are also questionable, and more research is needed.
Appetite management, body composition, and health
Though both sugar (227) and insulin (228) seem to suppress appetite/increase satiety, carbohydrate/sugar consumption might still promote hunger, food addiction, and excess energy intake.
As mentioned, (refined) carbohydrate/sugar-dense foods are often processed. Carbohydrate-dense processed foods may encourage overeating, given their hyper-palatability (sweet and/or savory), energy density, and consequent addictive addictive potential (192,196,197).
These factors encourage overconsumption, as hyper-palatability promotes greater food intake, energy density potentiates a calorie surplus, and addictive properties fuel this cycle. Practically, you might start eating cookies, struggle to stop eating them, consume many calories in the process, and want more in the future.
Further, the more processed the food, the less its consumption tends to blunt hunger. For example, one study (230) found that the most energy-dense processed breads actually satiated the least (per calorie). Thus, energy density negatively correlated with satiety index rating when subjects consumed isocaloric portions. This means that despite consuming just as many calories, the group that consumed the most energy-dense bread was likely to consume more throughout the day (since they’d be hungrier, possibly craving more energy-dense processed food).
While bread is a convenient example, many (aforementioned) other carbohydrate-ridden processed and addictive foods exist. Examples include: pizza, chocolate, ice cream, chips, cookies, french fries, cheeseburgers, cake, popcorn, muffins, soda, cereal, fried chicken, rolls, and gummy candy.
Notice, many of these highly processed food sources also contain fats. However, less energy-dense, processed carbohydrate sources with low-fat tend to satiate more per calorie (231).
For instance, french fries, lentils, pasta, rice, potatoes, fruits, fish, eggs, cheese, steak, beans, porridge, sustain, Special K, corn flakes, honey smacks, all-bran, popcorn, jelly beans, cookies, and crackers each tended to satiate more than white bread (231). While some of these foods contain processed carbohydrates (e.g. corn flakes, honey smacks, all-bran, Special K, pasta, popcorn, cookies, and french fries), they’re relatively satiating nonetheless. This indicates that some carbohydrate-dense foods can satiate well (especially porridge, potatoes, apples, and oranges) (231).
Meanwhile, croissants, cakes, donuts, mars bars, peanuts, potato chips, muesli, and ice cream each satiated less than white bread (on average) (231). This supports the assertion that carbohydrate-dense foods often induce less satiety per calorie. Since these processed carbohydrate sources often bear added fats, energy density, and low satiety index (231), it may be prudent to avoid consuming them. The degree to which any food satiates is unique to the individual, yet reducing processed carbohydrate intake typically promotes reduced calorie intake, less addictive food consumption, and suppressed hunger.
Generally, volume predicts a food’s influence on satiety to a greater degree than its energy content. This is evident as research indicates consuming satisfying portions of less energy-dense foods enhanced satiety, despite subjects’ reduced energy intake (232). Furthermore, greater food volume consistently satiates more than lower volume (iso-energetic) food (233–235). Therefore, swapping out energy-dense/processed carbohydrate sources for less energy-dense sources may help improve your body composition or health.
However, even if you only consume unprocessed, satiating, and non-addictive carbohydrate sources, your appetite can still benefit from reduced carbohydrate intake. Indeed, lower calorie diets (<800 kcals per day) improve satiety and reduce hunger (236), thus reducing your carbohydrate intake can improve satiety, insofar as it sufficiently reduces your energy intake. What’s more, cutting carbohydrate intake to <50 grams per day (ketogenic diet) can reduce hunger, improve satiety, and decrease desire to eat (236).
Additionally, swapping processed carbohydrates for lower-calorie vegetables and fruits can yield many health benefits (144–146,229,264–266). This is because processed foods also contain fewer micronutrients (210), whereas vegetables provide vitamins (149), minerals (149), and many other beneficial compounds (e.g. antioxidants (148,152), phytochemicals (150,152), and polyphenols (151)). Similarly, cutting processed carbohydrates (without replacement) can help improve body composition and health by reducing your energy intake, thus potentiating weight loss.
Note: None of this implies that a lower-carbohydrate diet yields superior fat loss; only that it might for very hungry individuals on a hypocaloric diet, via potentiating reduced calorie intake (as a function of reduced appetite).
This section neglects other appetite-influencing factors (e.g. protein (101,103,104), fiber (237,238), viscosity (239), and more discussed here (240)). Higher protein, fiber, viscosity, and ketone levels generally improve satiety/reduce appetite. Intermittent fasting (241), drinking fewer calories (239,242), and sleeping enough (243,244) may also help suppress hunger.
Taken together, this suggests that cutting your carbohydrate intake may improve your body composition. If you (over)consume many processed, energy-dense, addictive, less satiating, and/or hyper-palatable food sources, then switching to higher volume, less energy-dense, satiating sources should enhance health and body composition. Additionally, emphasizing fruit/vegetable intake can improve health, while boosting satiety and (likely) decreasing energy intake. Further, reducing calorie intake to <800 kcals per day (especially if emphasizing carbohydrate intake from vegetables) can reduce appetite and enhance satiety, potentiating decreased energy intake and fat loss. Lastly, if high desire to eat impedes your weight loss efforts (even when consuming <800 kcals/day), then reducing carbohydrate intake below 50 grams per day may reduce hunger, improve satiety, and decrease desire to eat.
Unfortunately, free sugar intake correlates closely with dental caries (cavities) (245–247). Indeed, nearly ¼ of U.S. adults have untreated dental caries, while ≤80% of people suffer cavities (245). Even “safe” free sugar intakes of ~50 grams/day (10% energy intake) induce “a costly burden of caries” (246,247).
Some authors recommend a pragmatic goal of ~3-5% energy intake (~15-30 grams) from free sugar to prevent adult caries. Currently, evidence suggests the “≤10% of energy intake recommendation” (~50 grams per day) produces many cavities (246,247), thus the pragmatic free sugar intake must be even lower. A 2016 review (245) affirmed that ≤10% of energy intake from free sugar, while faring better than >10% intake, does not eliminate caries. An intake of <5% of calories from free sugars should result in fewer cavities, or perhaps eliminate them.
However, note the distinction between “free sugars” and “sugars”. Free sugars include, “all monosaccharides and disaccharides added to foods…plus sugars naturally present in honey, syrups, and fruit juices”. This means that sugars from whole fruits, vegetables, nuts, seeds, legumes, and/or dairy do not count towards “free sugar” intake. Thus, you can still reap health benefits from fruits/vegetables with less concern, though non-free sugars may still threaten dental health.
While dental health is merely one subset of physical health, dental operations often yield pain, inconvenience, and financial burden. To minimize risk of dental caries, it’s prudent to limit free sugar intake.
Unfortunately, free sugar intake correlates closely with dental caries (cavities). Even “safe” free sugar intakes of ~50 grams/day (10% energy intake) can induce “a costly burden of caries”. A 2016 review affirmed that ≤10% of energy intake from free sugar, while faring better than >10% intake, does not eliminate caries. An intake of <5% of calories from free sugars should result in fewer caries, or perhaps eliminate them. However, note the distinction between “free sugars” and “sugars”. Free sugars include, “all monosaccharides and disaccharides added to foods…plus sugars naturally present in honey, syrups, and fruit juices”. This means that sugars from whole fruits, vegetables, nuts, seeds, legumes, and/or dairy do not count towards “free sugar” intake. To minimize cavity risk, it’s prudent to limit free sugar intake.
- The Carbohydrate Hypothesis of Obesity: a Critical Examination by Stephan Guyenet
- Bad sugar or bad journalism? An expert review of “The Case Against Sugar” by Stephan Guyenet
- The Case against Sugar Isn’t So Easily Dismissed by Gary Taubes, and the accompanying back-and-forth essay series with Stephan Guyenet and Yoni Freedhoff.
- Thoughts on a black swan 2 – the Hadza and honey by Gary Taubes
Taubes seemingly misconstrued the views of authors he cited in The Case Against Sugar (2016) publication. Quoting from Stephan Guyenet’s analysis:
“During the course of his argument, Taubes uses sleight of hand to portray the views of researchers as more favorable to his ideas than they really are. For example, in chapter 9 he argues that obesity and physical inactivity are not the real causes of insulin resistance, rather sugar causes both insulin resistance and obesity. To support his theory, he invokes the work of Stanford endocrinologist Gerald Reaven, claiming that he “was bringing back the notion that carbohydrates were bad”. This seemed mighty fishy to me, so I looked up what Reaven actually thinks. Here’s a quote from a review paper he wrote (emphasis mine) (32):
‘Since being overweight/obese and sedentary decreases insulin sensitivity, it is not surprising that the prevalence of the manifestations of the [insulin resistance syndrome] is increasing at a rapid rate. From a dietary standpoint, there are two approaches to attenuating the manifestations of the [insulin resistance syndrome]: (a) weight loss to enhance insulin sensitivity in those overweight/obese individuals who are insulin resistant/hyperinsulinemic; and (b) changes in macronutrient content of diets to avoid the adverse effects of the compensatory hyperinsulinemia [i.e., replacing carbohydrate with unsaturated fat- SG].’
Taubes neglects to inform the reader that Reaven thinks obesity and physical inactivity cause insulin resistance, and that these factors explain the rising prevalence of metabolic disease– precisely what Taubes is arguing against in that passage. Furthermore, Reaven explains in no uncertain terms that he does not think insulin resistance causes weight gain.
Rather than straightforwardly reporting what Reaven’s studies revealed and what the man believes, Taubes takes Reaven’s argument that people with existing insulin resistance may benefit from carbohydrate restriction and warps it to make it appear as if Reaven supports Taubes’s beliefs about the origins of insulin resistance. In doing so, Taubes flips Reaven’s position by 180 degrees. If you want the real scoop on Reaven’s important work, go straight to Reaven’s book Syndrome X.”
While I haven’t investigated all Taubes’ citations, it’s noteworthy that they don’t necessarily support his views.