How do I calculate the carbohydrates in my dog’s food?

Direct answer: use nitrogen-free extract by difference when the current label does not list carbohydrate, but treat the result as an estimate.

For most healthy dogs, the exact number is not decision-critical. The calculation is useful when you want to compare foods on the same calorie basis, check whether a food is broadly starch-forward, or give a veterinary nutritionist a transparent starting estimate. It is not a measure of starch quality, glycemic load, grain-free status, or overall food quality.

On most current U.S. labels, calculate:

NFE (% as fed) = 100 - (moisture + crude protein + crude fat + crude fiber + ash)

Then convert the result to an energy basis:

NFE (g/1,000 kcal) = NFE % as fed × 10,000 ÷ kcal/kg

Use the manufacturer's typical or average analysis when available. Guaranteed-analysis values are minimums and maximums, not expected averages, and ash is often missing. If ash is unavailable, approximately 6% for dry food and 2% for canned food are defensible temporary assumptions from Hill and colleagues, but they came from small measured subsets and must be stated as assumptions.

Nara is a canine health and longevity platform. We treat carbohydrate as one part of the diet's calorie allocation, not as an automatic problem. The useful question is not simply whether the number is low. It is whether the dog's full diet provides appropriate calories, protein, fat, fiber, nutrient adequacy, and consistency for that dog.

Evidence confidence: Moderate overall.

Confidence is high in the NFE arithmetic and in the distinction between current-label NFE and the modernized Total Carbohydrate declaration. Confidence is moderate in generic missing-ash assumptions and in any interpretation based only on guaranteed analysis. Confidence is low that small NFE differences predict glycemic control, disease risk, or food quality.

The answer in brief

Current U.S. dog-food labels usually do not require carbohydrate, so owners often estimate it as nitrogen-free extract after subtracting moisture, protein, fat, crude fiber, and ash from 100. The estimate should then be converted to grams per 1,000 kcal before foods with different moisture or calorie density are compared. Missing ash, guaranteed-analysis bounds, and crude fiber can each move the result, which is why typical analysis is preferable and small differences should not be ranked. The modern Pet Nutrition Facts format uses a different Total Carbohydrate calculation that includes dietary fiber and will also show grams per household unit and calories from carbohydrate. The two label systems should not be mixed.

A worked example first

Assume a dry food lists:

• moisture: 10%
• crude protein: 26%
• crude fat: 15%
• crude fiber: 4%
• ash: not listed, so 6% is stated as an assumption
• calorie density: 3,600 kcal/kg

The calculation is:

100 - (10 + 26 + 15 + 4 + 6) = 39% NFE as fed

Then:

39 × 10,000 ÷ 3,600 = 108 g NFE/1,000 kcal

That is the useful output: approximately 108 g of estimated NFE per 1,000 kcal, based on guaranteed-analysis values and an assumed 6% ash.

The table below shows why the assumptions matter.

A two-point ash difference moved the estimate by about 5.6 g/1,000 kcal. Applying Hill's average corrections to all five guaranteed-analysis inputs moved it by about 7.5 g/1,000 kcal. Neither shift proves that one food is meaningfully better than another. They show why label-derived carbohydrate should be reported as an estimate rather than a precise assay.

Current NFE and modern Total Carbohydrate are different numbers

This distinction is the most important part of the article.

The current-label estimate subtracts crude fiber:

NFE = 100 - (protein + fat + crude fiber + moisture + ash)

The modernized Pet Nutrition Facts calculation does not:

Total Carbohydrate = 100 - (protein + fat + moisture + ash)

The modern Total Carbohydrate value therefore includes dietary fiber. Do not compare a future bag's declared Total Carbohydrate with a crude-fiber NFE number as though they were produced by the same formula. Total Carbohydrate is also declared as a maximum guarantee, so it is the correct label-defined quantity for the new format, not an exact average recipe value.

AAFCO's model Pet Nutrition Facts box also adds information that is more useful than a percentage alone:

• the familiar household unit, such as a cup, can, treat, or piece, and its weight in grams;
• nutrient grams per that household unit;
• total calorie content; and
• calories from protein, fat, and carbohydrate.

That means a modern label can answer both "how many carbohydrate grams are in the amount I feed?" and "how much of the food's energy comes from carbohydrate?" without asking the owner to reconstruct both calculations.

AAFCO approved the revised model regulations in July 2023 and recommended a six-year enforcement-discretion period based on the printed 2024 Official Publication. Individual states control adoption and enforcement, so current and modernized labels can coexist during the transition. Check which format is actually on the package before calculating anything. [AAFCO PFLM Nutrition Facts; AAFCO Enforcement Recommendation; N.D. Admin. Code 7-17-13-01]

Why ash cannot simply be left out

Ash is the mineral residue left after a food is completely burned in laboratory analysis. It is not an added filler. It represents minerals and is part of the food's mass.

If ash is omitted from the subtraction, every missing ash percentage point is incorrectly counted as carbohydrate. A food with 7% ash would therefore have its NFE overstated by 7 percentage points if the calculation used zero ash.

Hill and colleagues compared guaranteed and measured composition across a large set of commercial pet-food inspection records. The frequently cited ash defaults did not come from the full dataset. Ash comparisons were available for only 39 analyses, with measured means of approximately 5.9% as fed in 12 dry foods and 2.2% in 22 canned foods. The authors rounded these to 6% and 2% for use when ash was unknown, while warning that foods with ash guarantees may have been selected partly to emphasize low ash. [Hill 2009]

An independent dataset gives useful corroboration for dry food. In 119 commercial extruded dry dog-food samples, mean ash was 6.25% as fed, with a standard deviation of 1.21 and a range from 3.25% to 10.24%. That makes 6% a reasonable central assumption for ordinary dry food, but the wide range confirms that it is not a universal value. The samples were supplied by a commercial manufacturer that also supported the work, and the study was designed around mineral prediction rather than market-representative carbohydrate estimation. It should be treated as corroboration, not as a definitive market average. [Goi 2019]

Use this hierarchy:

1. manufacturer typical or average ash;
2. a laboratory or complete nutrient analysis;
3. a clearly stated 6% dry or 2% canned assumption when no better value exists; or
4. "not estimable" when the food type makes a generic assumption indefensible.

For raw or bone-in diets, do not apply the 6% or 2% default. Bone can change ash substantially. Ask the manufacturer for a complete typical analysis or use a formulated laboratory analysis. The same caution applies to home-prepared diets. For fresh or gently cooked commercial food, ask the manufacturer rather than inventing a third universal default.

Guaranteed analysis can move every input at once

A guaranteed analysis is designed to establish regulatory bounds:

• protein and fat are usually minimums;
• crude fiber and moisture are usually maximums; and
• ash, when declared, is usually a maximum.

Those values can be useful, but they are not the same as the food's typical composition.

Hill reported average measured minus guaranteed differences of:

• crude protein: +1.5 percentage points;
• crude fat: +1.0;
• crude fiber: -0.7;
• moisture: -4.0; and
• ash: -0.5.

Applying those mean corrections to the worked example changes NFE from 39% to 41.7%. The increase can seem counterintuitive because measured protein and fat are higher, but the average four-point drop from the guaranteed moisture maximum is the largest term and more than offsets them.

This is a population correction, not permission to manufacture an exact typical analysis for one product. The more defensible conclusion is that the guaranteed-analysis result has an error band. Ask for typical values when the distinction matters. [Hill 2009; Tufts Petfoodology Nutrient Calculator]

Crude fiber is not total dietary fiber

Crude fiber is an older analytical fraction. It does not capture all soluble and fermentable fiber, so NFE calculated with crude fiber can count some fiber inside the carbohydrate remainder.

Traughber and colleagues analyzed 51 dog foods. Only a few failed to conform to their guarantees for the other measured macronutrients, while crude fiber was the clear exception: approximately 40% exceeded the guaranteed crude-fiber maximum. The authors described the crude-fiber and total-dietary-fiber relationship as moderate; the reported r = 0.843 is a strong numerical association in ordinary statistical language, but it still does not make the two measurements interchangeable. A systematic difference still moves the NFE residual. [Traughber 2021]

Earlier method-comparison work reached the same practical conclusion from a different direction. In a small set of dry dog and cat foods, crude fiber produced the lowest fiber values and did not correlate with the other fiber methods, while the total-dietary-fiber version of the NFE calculation was closest to measured starch. [de-Oliveira 2012]

This is why NFE is best described as a residual mass estimate. It can include digestible carbohydrate, fiber missed by the crude-fiber method, and accumulated analytical error. It is not measured starch or sugar.

Compare foods by calories, not by as-fed percentage

As-fed percentages are heavily influenced by water. A wet food can look dramatically lower in carbohydrate even when the difference is smaller once equal calories are compared.

For example:

• moisture 78%
• protein 8%
• fat 5%
• crude fiber 1%
• ash 2%
• 1,000 kcal/kg

The NFE estimate is:

100 - (78 + 8 + 5 + 1 + 2) = 6%

That looks far lower than the dry example's 39%. But on an energy basis:

6 × 10,000 ÷ 1,000 = 60 g/1,000 kcal

The useful comparison is therefore 60 versus 108 g/1,000 kcal, not 6% versus 39%.

You can also report grams per 100 kcal:

NFE (g/100 kcal) = NFE % as fed × 1,000 ÷ kcal/kg

Both energy-basis formulas are algebraically equivalent. Grams per 1,000 kcal are usually easier to read beside Nara's protein and fat calculations.

What counts as low, typical, or high?

There are no validated canine NFE cut-points that convert a label estimate into low, medium, high glycemic, or disease-risk categories.

A descriptive dry-food dataset can still provide context. Across 119 extruded dry dog-food samples, NFE averaged 42.98% as fed, with a standard deviation of 5.91 and a range from 23.70% to 54.05%. That tells us that a value around the low 40s is ordinary within that particular dry-food sample. It does not prove that 43% is optimal, that 24% is healthier, or that the same distribution applies to every market or food format. [Goi 2019]

Use broad comparisons only:

• a food near one end of a well-described format's distribution may reasonably be called lower or higher within that format;
• a difference such as 38% versus 42% should not be ranked without typical analyses and full dietary context; and
• wet, dry, fresh, and raw foods should not be placed into one as-fed percentage league table.

When the number matters, and when it does not

Healthy adult dogs

Adult non-reproducing dogs do not have a defined dietary carbohydrate requirement, but that does not make carbohydrate inherently harmful. Dogs can use cooked starch as an energy source, and a complete diet can be formulated across a wide range of macronutrient allocations.

For most healthy dogs, NFE should not decide between two complete foods by itself. Total calorie intake, body condition, muscle condition, protein intake, nutrient adequacy, digestive tolerance, and what the dog actually eats consistently are more useful.

Diabetic dogs

Do not manage canine diabetes from a home-calculated NFE number alone.

In a randomized crossover study of 12 dogs with stabilized spontaneous diabetes, Fleeman and colleagues compared two high-fiber, moderate-carbohydrate diets with a commercial moderate-fiber, lower-carbohydrate, higher-fat diet. There were no significant differences in insulin requirement or glycemic control among the diets. One high-fiber, moderate-carbohydrate diet produced weight loss and should not be routinely used in dogs with thin body condition. [Fleeman 2009]

The study does not prove that carbohydrate never matters. It shows that a coarse label residual is not a sufficient treatment rule. Calorie consistency, body condition, fiber characteristics, starch source, medication timing, clinical response, and the complete veterinary diet plan matter together.

Weight management

Energy density and total daily calories are the primary arithmetic. A higher-fiber food may support satiety for some dogs, but crude-fiber NFE becomes less reliable as an estimate of digestible carbohydrate precisely when fiber is substantial.

Grain-free or pulse-heavy diets

NFE cannot tell whether the carbohydrate came from rice, corn, potatoes, peas, lentils, tapioca, or soluble fiber. It cannot identify pulse concentration or resolve diet-associated dilated cardiomyopathy concerns. Ingredient pattern, formulation expertise, feeding trials, quality control, and manufacturer data are separate questions.

Therapeutic and home-prepared diets

Kidney, gastrointestinal, weight, and other therapeutic diets may be intentionally formulated around multiple interacting nutrient targets. Do not replace their full formulation with one NFE ranking. Home-prepared diets require a formulated nutrient analysis rather than subtraction from incomplete recipe percentages.

A transparent calculation workflow

1. Identify the label format. If the package has a Pet Nutrition Facts box with Total Carbohydrate and Dietary Fiber, use those declared values. Do not recalculate crude-fiber NFE and call it the same number.
2. Get the best inputs available. Ask for moisture, crude protein, crude fat, crude fiber or total dietary fiber, ash, and kcal/kg from a typical or average analysis.
3. State any ash assumption. If no typical ash is available, approximately 6% for conventional dry food or 2% for canned food can be used as an explicit estimate. Do not apply those defaults to raw or bone-in food.
4. Keep every value on an as-fed basis. Do not mix dry-matter values with as-fed label values in one subtraction.
5. Calculate NFE. 100 - moisture - protein - fat - crude fiber - ash.
6. Check the result. A negative result or a value that is obviously incompatible with the food means the inputs are inconsistent. Report "not estimable" rather than clipping it to zero.
7. Convert to an energy basis. NFE % × 10,000 ÷ kcal/kg gives g/1,000 kcal.
8. Name the provenance. Say whether the result used typical analysis, guaranteed analysis, or an ash assumption.
9. Interpret coarsely. Do not treat small differences as a ranking or a clinical threshold.
10. Add the rest of the diet. Treats, toppers, milk, chews, table food, oils, and other extras can change the day's intake more than a small difference between two bags.

For a broader whole-food check, see Does my dog’s food meet their calorie and protein needs?.

What this calculation cannot tell you

• It cannot measure starch or sugar directly.
• It cannot calculate glycemic index or glycemic load.
• It cannot identify the carbohydrate source.
• It cannot determine whether a food is complete and balanced.
• It cannot show amino-acid, fatty-acid, vitamin, or mineral adequacy.
• It cannot tell whether the diet fits the dog's calorie needs.
• It cannot compare wet and dry foods fairly until calories are equalized.
• It cannot support raw or bone-in estimates when ash is unknown.
• It cannot turn a three-point label difference into a meaningful health difference.

How Nara thinks about carbohydrate

Nara's interpretation

Nara treats carbohydrate as an allocation variable, not as an automatic negative.

When a product supplies a modern Total Carbohydrate declaration, Nara should preserve that value and its label-defined provenance. When only a current guaranteed analysis is available, Nara should calculate NFE by difference only when the necessary inputs can be supported, state any ash assumption, and convert the result to grams per 1,000 kcal.

The interpretation should then ask:

• how much of the day's energy carbohydrate occupies;
• whether protein and essential nutrients remain well supported;
• what fiber and ingredient context is present;
• what would replace carbohydrate if it were reduced;
• whether calorie intake fits the dog's body condition, muscle condition, and observed response; and
• whether the estimate covers the whole diet or only some components.

Nara would not use NFE as a food-quality score, a grain-free score, a diabetes target, or a reason to demonize a complete diet. A transparent estimate can be useful. False precision is not.

Evidence limitations

The formula itself is straightforward. The uncertainty comes from its inputs and from what owners may infer from the output.

• Hill's 6% dry and 2% canned ash values came from small subgroups, not from the full 2,208-food comparison.
• Hill's mean guaranteed-analysis corrections describe groups of foods and do not reveal one product's actual composition.
• The 119-food ash and NFE distribution came from one commercial sample source and was not designed as a representative survey of every food on the market.
• Crude fiber and total dietary fiber are analytically different, and neither makes NFE a direct starch assay.
• No validated canine minimum important difference tells us when two NFE estimates are meaningfully different.
• The canine diabetes trial was small and tested complete diets, not NFE values in isolation.
• The U.S. label transition depends on state adoption, so package formats will not change everywhere at once.

Sources

1. Association of American Feed Control Officials. Nutritional Labeling. https://www.aafco.org/resources/startups/nutritional-labeling/
2. Association of American Feed Control Officials. Calorie Content. https://www.aafco.org/resources/startups/calorie-content/
3. Association of American Feed Control Officials. Pet and Specialty Pet Nutrition Facts. July 2023. https://www.aafco.org/wp-content/uploads/2023/10/AAFCO-PFLM-Nutrition-Facts-Final-July-2023.pdf
4. Association of American Feed Control Officials. Position on Enforcement Discretion Related to Pet Food Label Modernization Changes. July 31, 2023. https://www.aafco.org/wp-content/uploads/2024/02/AAFCO-PFLM-Enforcement-Recommendation-Letter-FINAL.pdf
5. North Dakota Administrative Code 7-17-13-01. Pet and specialty pet nutrition facts. https://www.law.cornell.edu/regulations/north-dakota/N-D-A-C-7-17-13-01
6. Heinze CR. Carb Confusion: Part 2, Measuring and Comparing Carbohydrate in Pet Foods. Tufts Petfoodology. 2021. https://sites.tufts.edu/petfoodology/2021/09/27/carb-confusion-part-2-measuring-carbs/
7. Heinze CR. Carb Confusion Part 1: The Role of Carbohydrate in Pet Foods. Tufts Petfoodology. 2021. https://sites.tufts.edu/petfoodology/2021/07/07/the-role-of-carbohydrate-in-pet-foods/
8. Heinze CR. All About Ash. Tufts Petfoodology. 2022. https://sites.tufts.edu/petfoodology/2022/08/17/all-about-ash/
9. Clinical Nutrition Team, Cummings School of Veterinary Medicine. Pet Food Nutrient Calculator. Tufts Petfoodology. 2017. https://sites.tufts.edu/petfoodology/2017/08/07/nutrient_converter/
10. Hill RC, Choate CJ, Scott KC, Molenberghs G. Comparison of the guaranteed analysis with the measured nutrient composition of commercial pet foods. Journal of the American Veterinary Medical Association. 2009;234(3):347-351. https://pubmed.ncbi.nlm.nih.gov/19210254/
11. Traughber ZT, Detweiler KB, Price AK, et al. Effect of crude fiber and total dietary fiber on the calculated nitrogen-free extract and metabolizable energy content of various dog foods fed to client-owned dogs with osteoarthritis. American Journal of Veterinary Research. 2021;82(10):787-794. https://pubmed.ncbi.nlm.nih.gov/34554875/
12. Goi A, Simoni M, Righi F, et al. Prediction of mineral composition in commercial extruded dry dog food by near-infrared reflectance spectroscopy. Animals. 2019;9(9):640. https://pmc.ncbi.nlm.nih.gov/articles/PMC6770719/
13. de-Oliveira LD, Takakura FS, Kienzle E, et al. Fibre analysis and fibre digestibility in pet foods: a comparison of total dietary fibre, neutral and acid detergent fibre and crude fibre. Journal of Animal Physiology and Animal Nutrition. 2012;96(5):895-906. https://pubmed.ncbi.nlm.nih.gov/21812839/
14. Fleeman LM, Rand JS, Markwell PJ. Lack of advantage of high-fibre, moderate-carbohydrate diets in dogs with stabilised diabetes. Journal of Small Animal Practice. 2009;50(11):604-614. https://pubmed.ncbi.nlm.nih.gov/19814767/

Related questions

• What is the difference between Total Carbohydrate and NFE?
• Should dry and wet dog foods be compared by as-fed carbohydrate percentage?
• How much can missing ash change the estimate?
• Is lower carbohydrate automatically better for a healthy dog?
• Can NFE guide a diabetic dog's diet?
• Can NFE identify a grain-free or pulse-heavy food?

Get a dog-specific interpretation

Nara can turn a specific food label into a transparent carbohydrate estimate, show the ash and analysis assumptions, convert it to grams per 1,000 kcal, and interpret it beside the rest of the dog's recorded diet over time.