The Whole-Food Digestion Burn Came From One Lopsided Sandwich Test

What the answer rests on. Four single-meal tests carry the comparison: one that matched the meals for calories only (Barr and Wright, 2010) and three that also matched protein, carbohydrate and fat (Dioneda and colleagues, 2020; Mohr and colleagues, 2020; Galdino-Silva and colleagues, 2024). Two of the three matched tests come from one laboratory at Skidmore College, with overlapping authors, so they are not independent replications of each other; the Brazilian test is the independent one. A 4-week inpatient crossover trial (Hall and colleagues, 2019) adds whole-day data, a 6-week feeding trial (Karl and colleagues, 2017) adds a whole-grain diet, and a 2024 meta-analysis (Guarneiri and colleagues) adds protein.

The famous test (Barr and Wright, 2010, Food & Nutrition Research). 18 people were recruited (15 college students aged 18 to 22 and 3 adults aged 47 to 56); one 20-year-old man was excluded after an illness, leaving 12 women and 5 men. Each ate a whole-food meal (multi-grain bread with whole sunflower seeds and whole-grain kernels, and cheddar) and a processed meal (white bread and a processed cheese product, Kraft Singles in the appendix), as 600 or 800 kcal portions, on two mornings no more than a week apart. Oxygen use was sampled for 2 minutes each hour, and the respiratory quotient was estimated from the meals' composition rather than measured; total post-meal energy expenditure is the area under the response curve above the pre-meal rate, which the paper calls BMR, with 13 people measured for 6 hours and 4 for 5 hours. It came to 576.3 against 306.9 kJ (P = 0.0009), 19.9% against 10.7% of the meal's energy (P = 0.005), and the response lasted 5.8 against 4.8 hours. The meals differed in composition (protein 20% against 15%, carbohydrate 40% against 50%, fat 39% against 33%, about three times the fibre), which the abstract calls "comparable". Funding: the Howard Hughes Medical Institute and Pomona College.

The matched grilled-cheese test (Dioneda and colleagues, 2020, Nutrients). 11 healthy young women (12 recruited) each ate three meals in a single-blind randomized crossover: whole food (Ezekiel bread, Kate’s of Maine butter, Cabot Vermont Cheddar, Bolthouse fruit juice), processed (Wonder bread, I Can’t Believe It’s Not Butter, Classic American Kraft Singles, Fanta orange soda) and gluten-free. The authors describe the meals as matched for calories and macronutrients; their Table 1 lists 580 and 587 kcal, 67 g carbohydrate each, 26.5 and 28.5 g fat, 17.6 and 16 g protein, and 6 and 2 g fibre for the whole-food and processed meals (Table 3 gives 587 kcal for the gluten-free meal instead, a labelling slip in one of the two tables). The thermic effect was measured at 45 to 60, 105 to 120 and 165 to 180 minutes: 42.32 kcal (95% CI 31.85 to 52.78) after the whole-food meal and 36.32 kcal (25.86 to 46.78) after the processed meal, which the authors call similar; any gap between the two is FitChef's arithmetic, not a test the paper ran. The gluten-free version of the meal produced a lower response than both, which the authors report without pinning down a reason; this page draws no conclusion about gluten-free eating. Registered as NCT04440826; no external funding.

The independent breakfast test (Galdino-Silva and colleagues, 2024, Nutrients). A randomized parallel trial at the Federal University of Alagoas in Brazil: 42 adults with obesity were included (71.4% women; mean ages 28.6 and 29.9 years in the two groups) and 40 were analysed for post-meal energy expenditure. One group ate a breakfast of ultra-processed foods under the NOVA classification (commercial toast with cheese, ham, strawberry jam and margarine, and commercial guava juice with a fibre supplement), the other a similar breakfast without them (bakery bread toast with olive oil, fried eggs, and guava juice with honey), at 557 and 559 kcal with 8.5 g fibre each. Expenditure was read once, 75 minutes after the meal; it rose in both groups, and the group-by-time difference was not significant (p = 0.21). The authors say the design could not be double-blind, that the study was underpowered, and that the high similarity between the two meals (bakery bread against industrial toast as the carbohydrate base) may explain the result. Its Table 5 energy values are single readings expressed per day, not calories spent digesting the meal. The same breakfast test found the ultra-processed meal eaten faster (7:52 against 11:07 minutes) with fewer chews, a separate result about intake. Funding: CNPq and FAPEAL, with CAPES scholarships, none of which had a role in the design, analysis or writing; no conflicts declared.

The shake-and-bar test (Mohr and colleagues, 2020, Nutrients). 8 young women aged 18 to 24 (10 recruited) ate two breakfasts in a single-blind randomized crossover with a one-week washout: a whole-food breakfast (2% milk, Kashi Go-Lean Original cereal, almonds, raspberries and strawberries) and a meal replacement (Boost High-Protein Nutritional Drink, canola oil, honey and a Met-Rx meal bar). Both had 529 kcal, 69 g carbohydrate, 16 g fat and 32 g protein; sugar was 27.6 against 50.7 g and fibre 19.7 against 1.7 g. Over 120 minutes the thermic effect was greater after the meal replacement by 0.053 kcal per minute (p = 0.048), about 6 kcal across the two hours by FitChef's arithmetic, with no significant difference at individual time points (p = 0.845). The authors had hypothesized the opposite. No external funding.

Why the calorie figures do not line up. Each test measured over a different window: 2 hours (Mohr), 3 hours in three windows (Dioneda), 5 to 6 hours with hourly samples (Barr and Wright), a single reading at 75 minutes (Galdino-Silva), and 24 hours in a respiratory chamber plus two weeks of doubly labelled water (Hall). Their kcal figures are not comparable with one another and are not lined up on this page.

The whole-day data (Hall and colleagues, 2019, Cell Metabolism). 20 weight-stable adults at the NIH Clinical Center were randomized to an ultra-processed or an unprocessed diet for 2 weeks, then the other diet for 2 weeks, with meals matched for presented calories, energy density, macronutrients, sugar, sodium and fibre and eaten as freely as they wished. Intake was 508 ± 106 kcal/day higher on the ultra-processed diet (p = 0.0001). On weekly chamber days, with no significant difference in intake, 24-hour expenditure tended to be higher on the ultra-processed diet (51 ± 27 kcal/day; p = 0.06), which the authors contrast with Barr and Wright. Free-living expenditure by doubly labelled water was also higher on the ultra-processed diet (171 ± 56 kcal/day; p = 0.006), which the authors say was likely due to the differing states of energy balance between the diets, with intake higher on the ultra-processed diet. Funding: NIH Intramural Research Program.

Whole grains over six weeks (Karl and colleagues, 2017, American Journal of Clinical Nutrition). 81 men and postmenopausal women aged 40 to 65 were randomized after a 2-week run-in to 6 weeks of a weight-maintaining diet rich in whole grains (207 ± 39 g whole grains and 40 ± 5 g fibre a day) or based on refined grains (0 g whole grains and 21 ± 3 g fibre a day; 16 against 8 g fibre per 1000 kcal), with all food provided. The whole-grain group had a 92 kcal/day larger net daily energy loss (95% CI 28 to 156; n = 76; P = 0.005), attributed primarily to greater energy excretion in stool. Against the refined-grain group, the change in fibre-adjusted stool energy was 57 ± 17 kcal/day higher (total stool energy 96 ± 18) and the change in resting metabolic rate 43 ± 25 kcal/day higher, with P = 0.09 before adjustment and 0.04 after; the resting-rate difference was not significant once non-adherent participants were excluded, and the authors ask for cautious interpretation. The two components are not added here. Funding: the Bell Institute of Health and Nutrition, General Mills Inc., which reviewed and approved the protocol and manuscript but had no role in data collection, analysis or interpretation, and the USDA Agricultural Research Service; one author was employed by the Bell Institute during the study and another held an ASN 2012 Kraft Foods Inc. predoctoral fellowship.

Protein (Guarneiri and colleagues, 2024, Advances in Nutrition). Across 33 comparisons from 28 studies of meals, with diet-induced thermogenesis measured over 2 to 36 hours, higher-protein meals raised it (SMD 0.45; 95% CI 0.26, 0.65; P < 0.001) with substantial heterogeneity (I2 = 97.7%). The effect was significant when the difference in percent energy from protein between groups was 19.6% or more (18 comparisons; SMD 0.601) but not below that (15 comparisons; SMD 0.296; P = 0.051). The famous sandwich meals differed by 5% of energy (20% against 15%). Funded by General Mills; two authors (KK and BG-J) are General Mills employees.

What these sources do not show. No test here matched every nutrient: fibre differed in the grilled-cheese test (6 against 2 g) and in the shake-and-bar test (19.7 against 1.7 g), sugar and form differed in the shake-and-bar test, and the Brazilian test matched fibre and salt but read expenditure once. Each test fed one meal of each kind, to 8 to 40 people, and none followed body weight. The whole-grain trial compares whole with refined grains, not whole foods with processed foods in general.