The Atwater Factors — Why Modern Science Adjusts Them

The Atwater factors — the 4 kcal/g for protein, 4 kcal/g for carbohydrate, and 9 kcal/g for fat that underpin every food label and nutrition database on Earth — are derived from metabolic ward experiments (see also: where the 4-4-9 rule breaks down in practice) conducted by Wilbur Olin Atwater between 1887 and 1902 using respiratory calorimeters, fecal and urine collection, and a subject pool that consisted almost entirely of young American men eating late-Victorian diets of bread, beans, and meat, a methodological provenance that modern nutritional scientists have been systematically revising and extending for over 50 years without achieving consensus on a universally adopted replacement system. The FAO/WHO 2003 joint expert consultation (“Food Energy — Methods of Analysis and Conversion Factors”) formally concluded that the general Atwater factors introduce systematic errors of 5–10 % for many modern food categories and recommended adoption of food-specific factors for protein, fat, and carbohydrate from different botanical and food-processing origins — a recommendation that the FDA partially implemented in 2016 for dietary fiber but has not fully extended to proteins or fats.

The Original Atwater Experiments: Method and Scope

Atwater’s experiments at Wesleyan University between 1887 and 1902 used a human respiration calorimeter — a sealed, insulated room large enough to house a human subject for multiple days — that simultaneously measured CO₂ production, O₂ consumption, and heat dissipation from the occupant’s body. This “direct plus indirect” calorimetry approach allowed Atwater to calculate both heat production (metabolic rate) and substrate oxidation rates. Food intake was weighed and chemically analyzed; fecal and urinary output was collected in total and combusted in bomb calorimeters to determine energy losses.¹

From these data, Atwater calculated digestibility coefficients for protein, fat, and carbohydrate in specific foods — the fraction of each macronutrient that was absorbed across the intestinal wall rather than excreted in feces. He then computed metabolizable energy by subtracting both fecal losses (undigested macronutrient) and urinary losses (metabolic end-products, primarily urea from protein catabolism) from gross energy measured by bomb calorimetry.

The critical limitation is the study scope: Atwater tested approximately 100 food items on small numbers of subjects (typically 2–4 per food), all consuming what he described as “the American diet” of the 1890s. He explicitly averaged across food groups to produce general coefficients, noting in his published reports that these were intended for use when food-specific data were unavailable. The 4/4/9 system — rounded from his calculated values of approximately 4.35 kcal/g protein, 4.1 kcal/g carbohydrate, and 9.4 kcal/g fat — was designed as a practical approximation for dietary assessment in populations, not as a precise measurement system for individual food products.¹

A century of regulatory adoption has treated these approximate population-average factors as precise physical constants. The gap between their design intent and their regulatory application is the source of the modern adjustment literature.

Protein: Digestibility-Corrected Amino Acid Score and True Ileal Digestibility

The general Atwater factor of 4 kcal/g for protein incorporates two corrections: an average digestibility of approximately 92 % (meaning 8 % of protein consumed is excreted in feces) and a urinary nitrogen correction of approximately 1.25 kcal/g (the energy lost in urinary urea from protein catabolism). The resulting metabolizable energy factor of ~4.35 kcal/g is then rounded to 4.0 kcal/g.

The problem is the 92 % average digestibility assumption. Modern protein quality assessment shows that protein digestibility varies substantially by source, ranging from 97–99 % for isolated whey and casein proteins to 85–88 % for wheat protein and pea protein. The PDCAAS (Protein Digestibility-Corrected Amino Acid Score) method, adopted by FAO in 1991, incorporated digestibility correction but used fecal digestibility — measuring protein that reached the colon — rather than true ileal digestibility, which measures protein absorbed before the colon.²

True ileal digestibility — the gold standard, measured by ileostomy studies or dual-isotope dilution methods — shows that colonic bacteria ferment some undigested protein, producing fermentation products that appear in feces as bacterial biomass rather than as undigested food protein. Fecal digestibility therefore overestimates true protein absorption by 3–5 percentage points. When this correction is applied, wheat protein ileal digestibility is approximately 86–88 %, pea protein 85–88 %, and soy protein 90–93 %, compared with dairy protein at 95–97 %.²

Applying these corrected digestibility values to the Atwater framework: wheat protein’s metabolizable energy should be approximately 3.78 kcal/g (as published in the FAO/WHO 2003 specific factors table), not 4.0 kcal/g. For a diet providing 100 g/day of wheat-based protein, the overestimation is approximately 22 kcal — small for any individual meal but systematic across a year of dietary analysis. The 2013 FAO adoption of DIAAS (Digestible Indispensable Amino Acid Score) as the preferred protein quality metric further refined this, incorporating amino acid-specific digestibility coefficients rather than a single protein digestibility value.²

Fat: Chain Length, Saturation, and Structural Position

Atwater assigned all dietary fat a uniform 9 kcal/g, but modern data show that the metabolizable energy of fat varies by fatty acid chain length, degree of saturation, and structural position on the triglyceride glycerol backbone.

Medium-chain triglycerides (MCTs, C8–C12) yield approximately 8.3 kcal/g metabolizable energy rather than 9 kcal/g. The deficit arises from their metabolic handling: MCTs are absorbed directly into the portal blood without incorporation into chylomicrons, and are preferentially oxidized in the liver via rapid beta-oxidation rather than stored as adipose tissue. The obligate hepatic partial oxidation and the thermogenic effect of this rapid oxidation account for approximately 0.7 kcal/g reduction in net energy availability.³

Structural position on the triglyceride also matters. Stearic acid (C18:0, a saturated fatty acid) at the sn-2 position — as found in cocoa butter and some dairy fats — has significantly higher intestinal absorption efficiency than stearic acid at the sn-1,3 positions. This is because pancreatic lipase preferentially cleaves fatty acids at the sn-1 and sn-3 positions, releasing free fatty acids that can form insoluble calcium soaps in the small intestine and reduce absorption. The net effect on metabolizable energy from cocoa-butter-rich diets is approximately 5–8 % lower than the Atwater factor would predict for equivalent total fat content.³

Conjugated linoleic acid (CLA) fractions in ruminant dairy fat (typically 0.5–1.5 % of milk fat) may provide 8.5–8.7 kcal/g rather than 9, due to partial non-digestibility and altered metabolic partitioning. For most practical dietary contexts — mixed diets with moderate fat intake from diverse sources — the aggregate error from applying 9 kcal/g uniformly to all fat is approximately 3–5 %, rising to 8–10 % for very-high-MCT diets or cocoa-butter-concentrated diets such as those used in ketogenic protocols.

The FAO/WHO 2003 Specific Factor System

The 2003 FAO/WHO joint expert consultation (“Food Energy — Methods of Analysis and Conversion Factors”) is the most comprehensive published revision of the Atwater system. The consultation reviewed 50+ years of post-Atwater metabolizable energy research and published a table of food-specific energy conversion factors for 10 macronutrient classes.⁴

Selected specific factors from the FAO/WHO 2003 table, compared with general Atwater factors:

Component	Atwater Factor (kcal/g)	FAO/WHO Specific Factor (kcal/g)
Animal protein	4.0	4.27
Wheat protein	4.0	3.78
Oat protein	4.0	3.70
Animal fat	9.0	9.02
Plant fat	9.0	8.84
Available carbohydrate	4.0	3.87 (by difference)
Soluble dietary fiber	0 (formerly 4.0)	2.0
Insoluble dietary fiber	0	0
Polyols	4.0	0.2–2.6 (by compound)

Applying the specific factor system to a typical European mixed diet reduces total calculated energy by approximately 3–5 % compared with the general 4/4/9 system. For specific food categories — high-fiber whole grains, legumes, nut-heavy diets — the reduction can reach 8–12 %. The practical effect for an individual calorie tracker eating 2,000 kcal/day by Atwater calculation is that their true metabolizable energy intake may be closer to 1,900–1,940 kcal/day — a difference of 60–100 kcal, or roughly the equivalent of one small apple per day.⁴

The 2016 FDA Fiber Update: Partial Implementation

The FDA’s 2016 Nutrition Facts Panel rule represented the most significant US-regulatory update to energy accounting since the original Atwater adoption. The new rule implemented two specific changes relevant to energy calculation:

First, the FDA adopted specific energy conversion factors for dietary fiber: 2 kcal/g for soluble/fermentable fiber and 0 kcal/g for insoluble fiber, replacing the previous implicit treatment of all fiber at 4 kcal/g under the carbohydrate-by-difference system. This change reduced the labeled calorie content of high-fiber foods — a 100-gram serving of a food with 10 g total fiber previously could show 4 × 10 = 40 kcal for that fiber; it now shows approximately 10–12 kcal for the same fiber if predominantly insoluble, or 20 kcal if predominantly soluble.⁵

Second, the FDA added polyol-specific energy factors, correcting the previous treatment of polyols (sugar alcohols) as equivalent to digestible carbohydrate. Erythritol was assigned 0 kcal/g; xylitol 2.4 kcal/g; sorbitol 2.6 kcal/g; mannitol 1.6 kcal/g.

However, the FDA did not update protein or fat factors to food-specific values, despite the FAO/WHO 2003 consultation’s recommendations. The result is a hybrid system: specific factors for fiber and polyols (where scientific evidence was most compelling and regulatory pressure most acute) combined with general factors (4/4/9) for protein, fat, and digestible carbohydrate — an acknowledged inconsistency that the research community has documented but that has not prompted further regulatory action through 2024.⁵

What Apps Should Do: Pragmatic Implementation

For a nutrition tracking application, the practical question is how to implement energy accounting accurately given that food databases contain macronutrient composition data but typically do not specify fatty acid chain length, protein source digestibility, or fiber solubility fractions for each item.

The pragmatic best practice is a three-tier approach. First, apply fiber-specific and polyol-specific energy factors where the database entry specifies these values — this is the FDA-mandated standard and is increasingly incorporated in USDA FoodData Central entries. Second, apply the general 4/4/9 factors for protein, fat, and digestible carbohydrate, accepting the systematic overestimation of 3–5 % for high-fiber or plant-protein-dominant diets. Third, document the energy conversion system used on the methodology page, including which fiber energy value is applied, so that users and researchers can interpret logged calorie counts in context.⁴

Claims of per-gram precision beyond ±5 % for mixed-dish energy estimates are not supportable regardless of which Atwater variant is applied, because portion weight estimation (±10–20 %) and within-food composition variability (±3–8 %) each introduce larger errors than the Atwater factor choice. The Atwater limitation is a second-order error against the first-order errors of portion estimation and database matching. The system is useful and accurate enough for behavioral feedback and trend monitoring; it is not precise enough for clinical metabolic accounting. That limitation predates the apps by about 120 years.

References

1. Merrill AL, Watt BK. Energy Value of Foods: Basis and Derivation. Agriculture Handbook No. 74. Washington, DC: USDA Agricultural Research Service, 1973.

2. FAO. Dietary Protein Quality Evaluation in Human Nutrition: Report of an FAO Expert Consultation. FAO Food and Nutrition Paper 92. Rome: FAO, 2013.

3. Seaton TB, Welle SL, Warenko MK, Campbell RG. “Thermic effect of medium-chain and long-chain triglycerides in man.” American Journal of Clinical Nutrition 44, no. 5 (1986): 630–634.

4. FAO/WHO. Food Energy — Methods of Analysis and Conversion Factors: Report of a Technical Workshop. FAO Food and Nutrition Paper 77. Rome: FAO, 2003.

5. U.S. Food and Drug Administration. “Final Rule: Food Labeling: Revision of the Nutrition and Supplement Facts Labels.” 81 FR 33742, May 27, 2016. https://www.federalregister.gov/documents/2016/05/27/2016-11867