How does obesity develop—and are specific foods to blame?

The question of whether certain foods are culpable in the development of obesity has long provoked intense interest and spirited debate. At its core lies the assumption that avoiding particular foods or dietary components might prevent or reverse obesity.

Over time, numerous hypotheses have emerged to explain the dietary drivers of obesity: excessive fat intake, insufficient protein, high consumption of free sugars or refined carbohydrates, and lately, the degree of food processing and even non-nutritive factors such as environmental contaminants that disrupt hormonal and metabolic homeostasis. Any putative factor promoting net fat deposition and adipose tissue expansion must operate through one or more biological mechanisms that affect energy intake, energy expenditure, or nutrient partitioning toward storage or oxidation (1,2).

Two dominant conceptual frameworks shape this scientific discourse: the energy balance model (EBM) and the carbohydrate-insulin model (CIM). The EBM attributes obesity to calorie-dense foods that promote overconsumption and sustained positive energy balance. In contrast, the CIM posits that refined, high-glycemic index (GI) carbohydrates trigger hormonal responses that favor nutrient storage over oxidation, reduce circulating fuels in the late postprandial period, and stimulate more rapid rebound hunger, than lower GI foods (1,2).

GI reflects a food’s capacity to raise blood glucose concentration relative to a reference food (typically glucose or white bread). To account for carbohydrate quantity, glycemic load (GL)—the product of GI and bioavailable carbohydrate per serving—is often used.

A recent parallel-group trial by Liu et al. in 120 healthy, normal-weight Chinese adults tested the effects of isocaloric breakfast meals (880 kcal for males and 650 kcal for females) differing only in GI/GL: low [≤55], moderate [55–70], and high [>70] (3). The macronutrient composition was standardized (60% carbohydrate, 20% fat, 20% protein) and corresponding fiber contents were 3.9, 5.5, and 1.3 g per 100 g of food, respectively. Hormonal and metabolic markers and hunger ratings were tracked over 5 hours, followed by an ad libitum lunch. A baseline ad libitum lunch on the previous day allowed calculation of within-group change scores.

While absolute energy intake at the test lunch did not differ between groups, participants in the medium- and high-GI groups consumed 130–145 kcal more than at the baseline lunch—findings broadly consistent with the CIM, though not definitive.

Curiously, self-reported hunger ratings over the 5-hour postprandial period were similar across groups. Although appetite ratings are highly variable between individuals, they are reproducible at the group level (4,5), and usually predict subsequent energy intake (6)—but not invariably (7). The increased food intake observed here was thus not associated with greater subjective hunger, challenging CIM predictions. A prior crossover study with similar design (ad libitum lunch 5 hours after consuming an isocaloric high-carbohydrate or high-fat breakfast) found consistently lower hunger ratings after the high-carbohydrate breakfast, but these were not accompanied by reduced energy intake at lunch (8). Both studies suggest that high-GL meals may promote subsequent “overeating” relative to the momentary state of hunger—possibly via differences in cognitive or sensory cues related to the test meals (9), or central hedonic reward mechanisms that encourage food-seeking behavior and eating (10), rather than the peripheral metabolic pathways proposed by CIM.

The high-GI breakfast in Liu et al. elicited a greater and more prolonged rise in glucose and insulin, with glucose levels reversing later—consistent with CIM—but never falling below baseline (3). Surprisingly, the insulin surge was accompanied by somewhat less suppression of plasma free fatty acid and β-hydroxybutyrate concentrations, resulting in increased rather than decreased total circulating fuel availability (quantified as the summed caloric equivalent of circulating glucose, lactate, free fatty acids, and β-hydroxybutyrate) after the high-GI meal, throughout the postprandial period. These metabolic shifts did not correlate with subsequent energy intake. Contrary to CIM’s framework, the cumulative insulin response to the breakfast meal (quantified as the area under the curve) was associated inversely with lunch intake (3), corroborating previous observations (11).

Ad libitum energy intake at a fixed-time lunch after a different breakfast (3,8) serves as a satiety marker for that breakfast. However, satiety—referring to the processes that elicit a feeling of fullness after eating—encompasses both the magnitude and the duration of hunger suppression (4,5). Offering appealing foods at fixed times, even before hunger peaks, can trigger eating via sensory and hedonic cues, while delaying access when hunger is high can alter both subjective appetite ratings and subsequent food intake (12,13).

In crossover studies where isocaloric high-carbohydrate or high-fat breakfast meals (485 kcal) (12) or lunch meals (840 kcal) (14) were consumed by normal-weight males and females, the high-carbohydrate meals prolonged satiety and delayed the time at which participants requested to eat the next meal by 50–70 minutes. In one of these studies, the high-carbohydrate breakfast (cornflakes) had a GI of ~66 and a fiber content of 1.0 g per 100 g food, while the high-fat breakfast (eggs and bacon) had a GI of ~45 and a fiber content of 2.9 g per 100 g food (12)—a contrast similar to the intervention meals by Liu and colleagues.

So where does this leave us? Depending on one’s interpretive lens, Liu et al.’s findings may be seen as supporting or refuting the CIM (15,16). In theory, testing the CIM and the EBM should be straightforward: provide free access to high- vs. low-carbohydrate/GI foods (CIM) (17,18), or to more calorie-dense vs. less calorie-dense foods (EBM) (19,20), and measure ad libitum intake alongside appetite ratings and metabolic profiling. In practice, ecological validity is a major hurdle, not least because of the subjectivity of self-reported appetite and the typically short duration of nutrition intervention studies relative to the timescale of obesity development.

Humans eat for many reasons besides meeting their energy and nutrient needs. Food choice and consumption are influenced by factors such as macronutrient composition, GI, physical form (e.g., solid vs. liquid), fiber content, energy density, sensory attributes and appeal (e.g., texture, palatability, variety), and environmental context (e.g., eating alone or with others, plate size) (4). Habitual eating patterns are also an important parameter. In Liu et al., for instance, stratification by habitual breakfast consumption revealed that GI effects on subsequent intake were evident only among breakfast-naive participants and not among those who regularly consumed breakfast. This suggests that adaptation to high-GI intakes (the Asian diet is typically a high carbohydrate, high-GI diet) may blunt acute effects, akin to the diminishing impact of more calorie-dense ultraprocessed foods vs. less calorie-dense minimally processed foods on energy intake over time (daily exposure for 2 weeks) (19).

The amount of food consumed (volume or weight) and, by extrapolation, its energy density, is another common confounder. Humans learn, through experience, about the satiating effects of food and can therefore estimate the appropriate amount needed of each food/meal to elicit satiation when presented to them ad libitum (5). Covert manipulations of energy density (increase or decrease) do not lead to compensatory changes in the amount of food consumed (decrease or increase, respectively), and therefore result in proportional changes in total energy intake—both under carefully controlled laboratory conditions and in real life settings (21-24).

Differences in energy density are frequently present in comparative evaluations of diets rich in fat vs. carbohydrate (17,18), or in ultraprocessed vs. minimally processed foods (19,20), complicating efforts to attribute changes in total energy intake to any single factor—be it macronutrient composition, degree of processing, or energy density itself. Efforts to match energy density across macronutrients can in turn lead to differences in meal volume/weight, potentially prompting sensory and cognitive differences that influence intake. In this context, the study design by Liu and colleagues—standardizing calorie and macronutrient intake across intervention meals and allowing only the GI to differ, while assessing satiety via ad libitum intake at the same test meal—represents a methodological advance. It would be informative to examine whether similar effects persist across repeated exposures to this paradigm over multiple days.

While Liu et al.’s findings are primarily mechanistic, they carry potential implications for dietary guidance. The modest increase in energy intake following high-GI breakfasts—despite unchanged hunger ratings—suggests that carbohydrate quality or sensory and cognitive associations between different test foods may influence eating behavior in certain individuals, particularly those unaccustomed to such foods. However, the effect size was small and unlikely to override more dominant influences such as overall dietary pattern, energy density, and the broader food environment. For researchers, these findings highlight the need to account for and control the many food attributes and covariates that can shape intake. For practitioners, they reinforce the value of whole-diet quality—emphasizing lower dietary energy density and fiber-rich carbohydrate sources—rather than targeting GI in isolation. Translating acute, single-meal effects into meaningful long-term strategies will require studies that capture habitual eating behaviors in free-living conditions over extended durations.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.