A subtraction performed on a subtraction#
Net carbs is arithmetic you do yourself. Take the Total Carbohydrate line on a package, remove the grams of dietary fiber, remove the grams of sugar alcohol, and treat what is left as the carbohydrate that actually reaches your bloodstream as glucose. The reasoning behind it is legitimate: fiber and polyols are carbohydrate your small intestine largely cannot convert into blood sugar, so counting them beside starch and sucrose overstates what a food does to you.
What almost nobody notices is what the subtraction is being performed on. Under US labeling law, "total carbohydrate content shall be calculated by subtraction of the sum of the crude protein, total fat, moisture, and ash from the total weight of the food"1. Nothing about carbohydrate is measured directly. It is the residue left when four other things have been weighed, which means every analytical error elsewhere in the panel lands in that one column. Net carbs then removes two more lines from that residue. And the term itself has no legal existence: the components permitted to appear beneath Total Carbohydrate are dietary fiber, total sugars, added sugars and sugar alcohols, and that list is complete. No US regulation contains the phrase.
So the question worth asking is not whether to use net carbs. It is which of the two subtractions holds up. One is a rough kitchen-table version of a quantity food scientists define carefully. The other collapses eight chemically distinct compounds into a single word, and they do not behave alike.
Food scientists call it available carbohydrate, and they refuse to lump it#
The idea underneath net carbs predates keto marketing by decades. Rather than sorting carbohydrate by molecule size or by what a label prints, carbohydrate chemists sort it by where it ends up. In their scheme, dietary carbohydrate divides into glycaemic carbohydrate, "digested and absorbed in the small intestine," and non-glycaemic carbohydrate, which "enter[s] the large intestine"3. The glycaemic fraction is then split further by how fast it is released — rapidly available glucose versus slowly available glucose — because two foods with identical available carbohydrate can deliver it over very different windows.
That is what a net-carb calculation is groping toward, and it is a reasonable thing to want. But notice what the professional version insists on that the folk version discards. The non-glycaemic side is not treated as one bucket: plant cell-wall polysaccharides, resistant starch and non-digestible oligosaccharides "should be measured and researched in their own right"3. The scientific scheme is a set of named fractions. Net carbs is a single deduction. That difference is where the errors come from, and it is the same reason the two families of fiber behave so differently despite sharing one line on the label.
The fiber term is roughly right#
Subtracting fiber is the defensible half of the formula, at least for a glycemic goal. Fiber, by definition, is not digested or absorbed in the small intestine, so it does not arrive as blood glucose, and leaving it inside a carbohydrate total you are using to predict a glucose response is straightforwardly wrong.
Two leaks are worth knowing about. Fiber that reaches the colon is fermented, and fermentation returns some energy to you — which is why fiber is not calorie-free even though it is glucose-free, a distinction worked through in the calorie values of the macros. And the fiber line on a US label can include resistant starch, which is starch that behaves like fiber in some foods, in some people, cooked one way and not another. Neither leak is large enough to abandon the subtraction. Both are large enough that "fiber = zero" is a convention rather than a fact. If you are going to track one carbohydrate number for health rather than for arithmetic, fiber itself is the better one.
The sugar-alcohol term is where the formula breaks#
Here the folk formula stops approximating and starts misinforming, because "sugar alcohol" is not a substance. It is a chemical family, and its members are not close relatives. A review of the polyols compiled their glycemic and insulinemic indices side by side, and the spread is enormous2. US labeling law, separately, assigns each one its own calorie value1. Put the two records in one table and the problem is visible in a glance:
| Sugar alcohol | Glycemic index | US label calorie value |
|---|---|---|
| Erythritol | 0 | 0 kcal/g |
| Mannitol | 0 | 1.6 kcal/g |
| Lactitol | 6 | 2.0 kcal/g |
| Isomalt | 9 | 2.0 kcal/g |
| Sorbitol | 9 | 2.6 kcal/g |
| Xylitol | 13 | 2.4 kcal/g |
| Maltitol | 35 | 2.1 kcal/g |
| Polyglycitol / hydrogenated starch hydrolysates | 39 | 3.0 kcal/g |
| Sucrose, for reference | 65 | 4 kcal/g |
Glycemic and insulinemic index values from Livesey, 2003; calorie values from 21 CFR 101.9(c)(1)(i), which names hydrogenated starch hydrolysates rather than polyglycitol specifically. Sucrose's index is Livesey's reference figure.
Subtract every gram of sugar alcohol and you have priced erythritol and polyglycitol identically. One is a zero on both counts. The other has a glycemic index of 39 and three calories in every gram.
Erythritol is the outlier that makes the whole subtraction look valid. It is absorbed in the small intestine but essentially not metabolized, leaving the body largely unchanged in urine — hence a glycemic index of 0 and a legislated calorie value of 0. Maltitol, the workhorse sweetener of sugar-free confectionery, is a different animal: a glycemic index of 35 puts it above lactitol, isomalt, sorbitol and xylitol combined in terms of glucose response, and it still carries 2.1 calories a gram. A product declaring 20 g of sugar alcohol tells you nothing about which of those two you just ate, and the gap between them is 39 index points and 3 calories per gram wide.
Europe already prints a net-carb figure — and it is the other one#
The transatlantic difference here is not a technicality; it inverts the formula. EU food law defines "carbohydrate" as "any carbohydrate which is metabolised by humans, and includes polyols," and defines "fibre" separately as "carbohydrate polymers with three or more monomeric units, which are neither digested nor absorbed in the human small intestine"4. The mandatory nutrition declaration lists "fat, saturates, carbohydrate, sugars, protein and salt"; polyols and fibre appear only in the optional list a manufacturer may add5.
Read that carefully and a European carbohydrate figure is already net of fiber and still inclusive of polyols — precisely the opposite pairing to the one a net-carb calculator wants. An American shopper subtracts fiber that is in the number; a European shopper subtracting fiber is removing grams that were never counted. This is one more branch of the same fork that makes two apps disagree about the same granola bar, and it is the single most common way people double-count their way to a wrong figure.
It is a blood-sugar estimate, not a calorie discount#
The most consequential misuse of net carbs has nothing to do with chemistry. People reach for it as a calorie exemption, and it is not one. Subtracting grams from a carbohydrate line does not remove energy from a food. Fiber still ferments. Polyols still run from 0 to 3 calories a gram. And whether the printed calorie figure has already discounted them is a manufacturer's choice you cannot see from the package, because US law offers several legal calculation methods that differ on exactly this point.
The cleanest way to hold both ideas at once is to keep them on separate ledgers. Total carbohydrate — the whole line, undiscounted — is what feeds your calorie arithmetic, since the calorie total is just the macro grams run through their conversion factors. Net carbs is a crude prediction of a glucose response, and predictions of glucose response are noisier than any chart implies: retesting white bread in the same person swung its glycemic index by more than it varies between different people. A number computed from a residual, minus a convention, minus a chemical family, is not a number to defend to the gram.
One footnote on the polyol most often subtracted. Plasma erythritol in the top quartile tracked roughly double the three-year risk of major adverse cardiovascular events across a US cohort of 2,149 and a European cohort of 833 (adjusted hazard ratios 1.80, 95% CI 1.18–2.77, and 2.21, 95% CI 1.20–4.07), with laboratory work showing enhanced platelet reactivity6. That is a real signal, and it is also contested on a specific, non-trivial ground: humans synthesize erythritol endogenously from glucose via the pentose phosphate pathway, no dietary information was collected, and elevated plasma levels may mark metabolic dysregulation rather than sweetener intake (Cramer et al., 2023, who declare no commercial interests). The two sides are not really in conflict — one measured an association, the other questions the exposure it was measuring. It is unresolved, and it belongs in the same sentence as the subtraction.
Counting it without fooling yourself#
Four rules follow from the definitions rather than from any diet philosophy.
Pick which question you are asking. For calories, use total carbohydrate and stop. For glucose management, subtract — the subtraction only ever answers the second question.
Subtract fiber in full, once. On a US label it is inside the total and comes out. On an EU label it was never in the total; subtracting it again is double-counting.
Do not subtract polyols wholesale. Erythritol comes out at 100% with a clear conscience. Maltitol and hydrogenated starch hydrolysates should come out at half or not at all — the widely used convention of halving polyols exists precisely because the family average is nothing like erythritol.
Use one rule, every day. Every convention above shifts a food's number by a few grams. A constant offset cancels out of the only comparison that matters, which is this week against last week. A rule you change midway does not.
FAQ#
How do you calculate net carbs?#
Total carbohydrate minus dietary fiber minus sugar alcohols, in grams. No US regulation defines the term or the method — the components permitted under Total Carbohydrate are fiber, total sugars, added sugars and sugar alcohols, and "net carbs" is not among them1. On an EU label, fiber is already excluded from the carbohydrate figure, so subtracting it again double-counts4.
Should sugar alcohols be subtracted from carbs?#
Not uniformly, because they are not one substance. Glycemic indices across the family run from 0 for erythritol and mannitol to 35 for maltitol and 39 for polyglycitol, and US label calorie values run from 0 to 3.0 kcal/g (Livesey, 2003; 21 CFR 101.9). Subtract erythritol fully; discount maltitol and hydrogenated starch hydrolysates by half at most.
Does counting net carbs lower a food's calories?#
No. The subtraction changes a carbohydrate figure, not an energy figure. Fiber is fermented in the colon and returns some energy, and polyols carry up to 3 calories a gram1. Whether those grams were already discounted in the printed calorie count depends on which permitted calculation method the manufacturer chose, which the package does not disclose.
Sources#
- 21 CFR 101.9 — Nutrition labeling of food: total carbohydrate calculation, permitted subcomponents, and specific caloric values for sugar alcohols. US Code of Federal Regulations (Cornell LII).
- Livesey G. Health potential of polyols as sugar replacers, with emphasis on low glycaemic properties. Nutr Res Rev. 2003;16(2):163-191.
- Englyst KN, Englyst HN. Carbohydrate bioavailability. Br J Nutr. 2005;94(1):1-11.
- Regulation (EU) No 1169/2011 on the provision of food information to consumers, Annex I — specific definitions of carbohydrate, sugars, polyols and fibre.
- Regulation (EU) No 1169/2011, Article 30 — content of the mandatory nutrition declaration and the optional supplementary list.
- Witkowski M, Nemet I, Alamri H, et al. The artificial sweetener erythritol and cardiovascular event risk. Nat Med. 2023;29(3):710-718.
- Cramer T, Gonder U, Kofler B. Plasma erythritol and cardiovascular risk: is there evidence for an association with dietary intake? Front Nutr. 2023;10:1195521.
