How is exogenous carbohydrate oxidation measured?

When a study reports that athletes oxidised 60 g of an ingested carbohydrate per hour, it sounds as simple as measuring heart rate. It is not. Measuring exogenous carbohydrate oxidation, a method sometimes referred to as ¹³C carbohydrate oxidation or carbon-13 breath tests, requires a deeper understanding of biochemistry, physiology, and stable isotope tracer methodology. Carbon-13 testing services are now also offered commercially, but they all use shortcuts to make them easier and more cost-effective. The services use stable isotopes, carbon-13-to-carbon-12 ratios in the breath, and total carbon dioxide production. However, the method works only if certain criteria are met and certain assumptions hold. If these criteria are not fully met or the assumptions are violated, the numbers can be wrong. This matters for the question of how much carbohydrate to take during exercise, especially because much of the evidence on oxidation at intakes above 90 g/h comes from this type of data (see Is more carbohydrate better?). It is therefore worth knowing what these methods, and the tests based on them, can and cannot tell you.

In this blog, we will look at what the carbon-13 method measures, which assumptions it makes, what a published critique of the most recent marathon study says, and what we can conclude about studies published in the scientific literature. The question of whether more carbohydrate improves performance is discussed in a separate blog (LINK).

Why do we measure the oxidation of ingested carbohydrate?

What you ingest and what you oxidise can be very different things. A drink can deliver 120 g of carbohydrate per hour, but if only 85 g of it is oxidised, the other 35 g is somewhere else. Most likely it is still in the stomach or intestine, or it may have been converted to glycogen or fat in the liver. Oxidation efficiency is the share of the ingested carbohydrate that is oxidised. If one ingests 100 g and 70 g is oxidised, the efficiency is 70%. Low efficiency means that a larger percentage of the ingested carbohydrate is not used as fuel and one hypothesis is that carbohydrate left in the gut contributes to gastrointestinal symptoms. Exogenous oxidation, the oxidation of the carbohydrate that was ingested, is also what showed us that a glucose-only drink cannot be oxidised faster than about 1 g/min, and that a mix of glucose and fructose can. The number has therefore been central to carbohydrate intake recommendations since the early 2000s.

What is carbon 13 testing?

Carbon-13, or ¹³C testing, in the context of endurance sports refers to the measurement of exogenous carbohydrate oxidation (the use of carbohydrate from your ingested drink or gel or other carbohydrate source). The testing uses the stable isotope ¹³C. What does that mean?

Carbon is a constituent of many molecules, fats, proteins, and carbohydrates (which even get their name from carbon); all consist mainly of carbon and hydrogen, with some oxygen and nitrogen in the case of protein. Carbon has an atomic mass of 12 (6 protons and 6 neutrons). But there is also a slightly heavier carbon, ¹³C (6 protons and 7 neutrons), and even ¹⁴C (6 protons and 8 neutrons). The atoms ¹²C and ¹³C are stable; they stay carbon 12 or carbon 13. We call them stable isotopes. ¹⁴C is not stable; it is radioactive. It falls apart and produces radiation (radioactive isotope). Of all the carbon in the atmosphere and around us, about 98.9% is ¹²C and about 1.1% is ¹³C.

High and low natural abundance of ¹³C

The 1.1% is an average. The exact value depends on how a plant fixes carbon dioxide during photosynthesis. Plants that use the C3 pathway (beet, potato, tapioca, wheat, rice) take up less ¹³C than plants that use the C4 pathway (maize, sugar cane). Papers report this as δ¹³C: the difference between the ¹³C/¹²C ratio of a sample and that of an international standard (VPDB), in parts per thousand (‰). Typical values are about −7 to −13‰ for C4 carbohydrate and about −19 to −27‰ for C3 carbohydrate. Table 1 converts these values into the ¹³C/¹²C ratio and into the percentage of all carbon that is ¹³C (my calculation).

Table 1. ¹³C/¹²C ratio and percentage of carbon that is ¹³C in different carbohydrate sources

The reason I list these numbers here is to make it clear that the differences between carbohydrates with a high and a low natural abundance are very small, but they matter for the interpretation of the results. A typical corn-derived carbohydrate contains about 1.094% ¹³C and a typical C3 carbohydrate about 1.077%: a difference of 0.016 percentage points, or about 1.5% in relative terms. A highly enriched ¹³C tracer is in a different league. Glucose with ¹³C on all six carbon atoms contains about 99% ¹³C, roughly 90 times the natural level. Mixed into a drink at 0.5 mg per gram of glucose, as in a recent study (1), it raises the drink to about 1.144% ¹³C. We can measure all these small differences with a method called isotope ratio mass spectrometry. These mass spectrometers can measure breath and also solids (drinks, gels), can cost hundreds of thousands of dollars and require skilled personnel to operate them.

Companies specialise in synthesising molecules that are close to 100% ¹³C, and even molecules with ¹³C labels at very specific positions. Researchers use these stable isotope tracers to gain better insight into specific pathways and metabolism in the body.

So in brief

• Some carbohydrates have a low natural abundance of ¹³C

• Some carbohydrates have a high natural abundance of ¹³C

• And there are tracers with a very high ¹³C content that can be used to enrich a molecule we want to study.

Because in most reactions in the body ¹³C and ¹²C react the same, and we can measure ¹³C/¹²C ratios, it is possible to use ¹³C as a label. If the ¹³C (or ¹²C) carbohydrate would be used preferentially in a biochemical reaction, we would call this an isotope effect.

How can stable isotope tracers be used?

The principle is easiest to see with an isotope other than carbon. Deuterium (²H) is a stable isotope of hydrogen with one extra neutron, and water made with it is known as heavy water. If a person drinks a known amount of deuterium-enriched water, the deuterium mixes with all the water in the body. After a few hours, a sample of saliva, urine or blood contains water with a lower deuterium concentration than the drink did, because the label has been diluted by the water already present. The more the label is diluted, the more water the body contains, so measuring the dilution allows total body water to be calculated.

A bucket of water makes a useful comparison. Imagine you need to know the volume of water in a bucket, but you cannot tip it out, and you cannot weigh it. You could add a known amount of dye, mix it thoroughly and measure the concentration of dye in the water. A low concentration means the dye has been spread through a large volume, and a high concentration means the volume is small. In this comparison the person is the bucket and deuterium is the dye. The body handles deuterium-labelled water in the same way as ordinary water, and at the doses used in research it is safe, so it can mark the body’s water without changing how the body behaves.

The same logic applies to nutrients. When a carbohydrate labelled with ¹³C is ingested, the label can be followed into the blood and into the carbon dioxide in the breath, which shows how much of the carbohydrate has been used for energy.

How can the carbon-13 be used to measure exogenous carbohydrate oxidation?

In 1977 Pirnay and colleagues applied these principles during exercise and gave volunteers glucose from a naturally carbon-13 rich source during prolonged exercise (2). When the glucose is oxidised, some of the carbon-13 ends up in the carbon dioxide in the breath, and the more carbon-13 in the breath, the more of the drink was oxidised.

Later studies used two approaches. Some used drinks made from corn-derived carbohydrate with a high natural carbon-13 content, as we did in some of our studies (3). In other studies a highly enriched tracer, such as glucose labelled with carbon-13 on every carbon atom, was added to the drink. In both cases the researcher measures the carbon-13 enrichment of the breath and the total volume of carbon dioxide produced. The calculation, in words, is:

Exogenous carbohydrate oxidation (g/min) = carbon dioxide production × (breath enrichment during the trial minus background breath enrichment) divided by (enrichment of the drink minus background enrichment) divided by 0.7467.

The number 0.7467 is the litres of carbon dioxide produced when one gram of glucose is oxidised. The calculation is sound, but only if every term in it is the right number. A review of gas exchange methods discusses the conditions for this in detail (4).

Why do such small differences matter?

The signal the method detects is small, and the calculation magnifies small errors in it. A difference of 1‰ in δ¹³C is a difference of 0.1% in the ¹³C/¹²C ratio, or 0.0011 percentage points of all carbon (my calculation). Yet this is the scale on which the result is decided. The whole difference between a typical C3 and a typical C4 carbohydrate is about 15‰, and the rise in breath δ¹³C caused by a maize-derived drink is only a few ‰.

A worked example shows how quickly this adds up. Take an athlete producing 3.0 L/min of CO₂ and oxidising 0.90 g/min of the drink, with a true background of −23.0‰. With a corn-derived drink (−11‰) the breath δ¹³C rises by only 2.7‰ above the background. With a drink enriched with a tracer (+34.7‰, as in the Birmingham study (1)) the same oxidation raises it by 12.9‰. Table 2 shows what happens if the background used in the calculation is lower than the true one.

Table 2. Effect of an error in the assumed background on calculated exogenous oxidation (illustration, my calculation)

Assumed background lower than the true background by	Maize-derived drink (−11‰): calculated oxidation (error)	Enriched drink (+34.7‰): calculated oxidation (error)
0‰ (correct background)	0.90 g/min (0%)	0.90 g/min (0%)
0.5‰	1.03 g/min (+13.9%)	0.93 g/min (+3.0%)
1‰	1.14 g/min (+26.6%)	0.95 g/min (+5.9%)
2‰	1.35 g/min (+49.5%)	1.00 g/min (+11.6%)

The values are an illustration, not data from a study.

An error of 1‰ in the background therefore makes a maize-derived drink look 27% more oxidised, and the enriched drink 6%. The enriched drink is more forgiving because its signal is larger relative to the error, which is the argument made for tracers (5). Other errors add to this. The same 1‰ in the δ¹³C of the drink changes the denominator of the equation by about 9% (12‰ instead of 11‰ in the example), and an error in VCO₂ changes the result in proportion.

This is how small errors turn into wrong conclusions. At an intake of 90 g/h, the true oxidation of 0.90 g/min is an efficiency of 60%. With a 1‰ background error and a maize-derived drink, the calculated efficiency is 76%. That is larger than the difference in efficiency between 90 and 120 g/h in an independent laboratory (86% versus 76%) (6). If the background error differs between two conditions, for example because of different intensities or diets, it creates a difference between conditions that does not exist. This is why the background is one of the central issues in the rest of this blog. But there are more potential methodological traps.

Which assumptions does the method make

There are several assumptions we make when using this methodology. If one of those assumptions is not true in the situation where we measure it, the numbers produced will be inaccurate or even wrong.

• The tracer behaves like the carbohydrate that is traced. If carbon-13 labelled glucose is used to trace maltodextrin, the two have to be oxidised at the same rate. Personally, I think that there may not be a problem with using a glucose tracer to trace maltodextrin based on the fact that maltodextrin oxidation in studies has produced similar results to glucose. Having said that, no one has ever done a direct comparison in one study. The point is still a really important one in general though and the tracer-tracee matching is a really important issue, especially in studies where dual tracers are used, such as in studies that have investigated glucose:fructose mixes. The infographic below is from a study that illustrates this quite well.

• The carbon-13 in the breath comes from the drink and not from carbohydrate stored in the body. More accurately: the background is the breath enrichment that would be present during exercise without the drink. If glycogen is high in ¹³C and we use more glycogen during exercise, the ¹³C in the breath rises independently of the ingested carbohydrate. This is the background problem we will discuss in more detail below.

• All CO₂ leaving the lungs comes from the oxidation of carbohydrate and fat. This is a standard assumption for all indirect calorimetry measurements. When bicarbonate is given, or ketones are ingested, this assumption is violated and calculations of substrate use are no longer valid. It is also affected when the body releases carbon dioxide from its bicarbonate stores, for example at high intensities when anaerobic glycolysis is activated and lactate concentrations start to rise.

• All carbon-13 from the drink appears in the breath. This is not true during the early part of exercise, because part of the carbon-13 is temporarily trapped in the body bicarbonate pool. Only once this pool has been saturated does the appearance of ¹³C in the breath reflect oxidation. Estimates of how long this takes vary from about 30 minutes to about an hour, which is one reason why many studies use only the last part of a long trial. The pool is also disturbed by changes in intensity and by stops.

In this study (7) we asked cyclists to ride for 2.5 hours at a moderate intensity whilst ingesting one of 4 drinks. The first one was just water, purely for background corrections. One drink contained soluble starch (amylopectin) and one contained insoluble starch (amylose). This was the main comparison of the study: what are the differences in oxidation rates between soluble and insoluble starch? Another question in this study was whether we could use a glucose tracer to trace insoluble starch. The ¹³C-glucose tracer was added to the amylose drink and the oxidation was calculated. The differences, as shown in the B-section of the graphic were very large. The orange line is the oxidation rate using the natural enrichment of the amylose. The values are corrected for background and as accurate as they can be. The yellow line are the calculated values when a ¹³C glucose tracer was added. Adding the tracer clearly resulted in a large overestimation of the actual oxidation rates. This indicates that the assumption that glucose and amylose behave similarly was violated. The glucose was oxidised faster than the amylose and did not behave the same. If tracer and tracee do not behave the same, the results cannot be interpreted.

Tracer and tracee don’t behave the same

Another situation where tracer and tracee don’t match may occur when we study carbohydrate mixtures. For example, when studying glucose:fructose mixes, two tracers need to be added. If the ratios of the glucose tracer to tracee ratio and the fructose tracer to tracee ratio are not exactly the same, the ¹³C appearance may not reflect the true glucose:fructose oxidation. We know that fructose is oxidised slower than glucose and if the enrichments are not exactly the same and we are only measuring the total ¹³C, we cannot control for these different oxidation rates.

Unfortunately, in many studies, it is just reported that glucose and fructose tracers were added, but there is no measurement of the actual ¹³C:¹²C ratio of the enriched glucose and fructose before mixing, so we will never know if the enrichments were the same. As a result we will never know whether the values are an overestimation, an underestimation or the true values.

Why does the background matter so much?

One of the main assumptions is that the ¹³C in expired CO₂ is coming from the drink and not from other sources (i.e. glycogen) because this is how we can differentiate between carbohydrate from exogenous and endogenous sources. To achieve this, glycogen should be depleted several days before the test, reducing stores as much as possible, followed by a diet that avoids foods high in ¹³C. The trick is to avoid all foods that are corn or cane sugar derived (including many sauces and hidden ¹³C sources).

If glycogen stores contain ¹³C, this will not be visible at rest. However, during exercise, when glycogen stores are being used more and more, the ¹³C will start to appear in expired breath. If we conducted the exact same exercise session, but ingested water or a carbohydrate with a low ¹³C abundance, we would see a rise in ¹³C that would be from glycogen only. This is the background correction we need to make to make sure that we only study exogenous ¹³CO₂ production and not a combination of exogenous and endogenous sources. Taking a resting background sample is not good enough, as the background can shift significantly even when glycogen stores are depleted and a diet is followed. When glycogen depletion is not part of the protocol the background shift will be substantially larger (and several recent studies do not mention glycogen depletion in the methods section).

Preparation before the trial and background determination in three sets of studies

Study	Hard exercise before the trial	Diet before the trial	Background
Earlier studies from our laboratory (for example (3)	Glycogen-depleting exercise one week before	Diet free of C4 plants and C4-derived foods after the depletion	Water trial
Hearris et al. 2022 (9)	No glycogen depletion described. Incremental test to exhaustion at least 7 days before	2 days of minimising foods with a high natural abundance of 13C. 24 h at 8 g/kg carbohydrate without exercise	Water trial
Ravikanti et al. 2025 (8)	No glycogen depletion described. Incremental test at least 1 week before, then a 60-minute familiarisation run	24 h standardised diet at 8 g/kg carbohydrate. No restriction of C4 foods described	Resting breath sample only

There are two ways to correct for the background that remains. The approach of earlier studies, including ours, was to include a trial with water, so that the change in breath enrichment from rest to exercise could be measured in the same person on the same diet (3). Another approach is to use a trial with carbohydrate that is low in carbon-13 for the same purpose (6). Both are designed to answer the same question, namely how the background changes during exercise without the labelled drink.

“A resting breath sample does not tell us what the breath looks like during exercise without the tracer.”

Podlogar and Rowlands, in a published letter on the marathon study of Ravikanti and colleagues, state that the background in those studies was breath collected at rest (8,9). Substrate use and the respiratory exchange ratio change markedly from rest to heavy exercise, and a resting background can inflate the estimate. They cite work showing 13% higher estimated oxidation of a carbohydrate from maize when water was used as the comparison (8).

Even when the diet is controlled this can be a significant shift, and it varies between individuals, which could also explain at least in part the large variation between participants that is reported in some studies. Individuals have different diets, some will eat more corn derived products than others, some will use more cane sugar products. This was one of the reasons that no studies could be conducted in the US, because the vast majority of carbohydrate sources are from corn and cane sugar. In continental Europe beet, potato and wheat are the main sources of carbohydrate.

Limitations of indirect calorimetry

With indirect calorimetry we measure pulmonary gas exchange (V̇O₂ and V̇CO₂) at the mouth. The assumption we make is that what we measure at the mouth reflects what happens at the cellular level in the tissues. We measure at the mouth and draw conclusions about carbohydrate and fat oxidation in the tissues. There are many situations where this assumption is reasonable, but there are also situations where what we measure at the mouth does not reflect what happens at the tissue level. One example is high-intensity exercise, when hyperventilation causes excess CO₂ production that is unrelated to oxidative processes. The stoichiometric equations we use to calculate carbohydrate and fat oxidation assume that these are the only two substrates oxidised. In some situations protein can play a role, although this may not be more than about 5% in extreme situations. If other processes such as gluconeogenesis, lipogenesis or ketogenesis are activated, this also changes the VCO₂ to VO₂ ratio, and what we measure then does not reflect carbohydrate or fat oxidation accurately. An understanding of when this method can and cannot be used is therefore essential.

The marathon study

Going back to the letter to the editor from Podlogar and Rowlands about the marathon paper. The letter points out a number of potential issues.

The first is the tracer. Carbon-13 glucose was used to trace maltodextrin, whose origin and chain length were not reported. The authors of the marathon paper acknowledge that using glucose to trace maltodextrin needs direct validation (9).

The second is the steady state. The runners stopped for one minute at several points to be weighed, and blood lactate was not in a steady state. Both change how much carbon dioxide leaves the bicarbonate pool, which adds to the measured carbon dioxide production and can lead to an overestimate of carbohydrate oxidation.

The third is the design. In the marathon study the 60 g/h drink had no fructose, and fructose was added to reach 90 and 120 g/h, so the effect of dose cannot be separated from the effect of fructose (9). A comparison in which dose and composition change together answers a different question from the one it seems to answer.

Then there is the background. A resting sample was used, which is expected to overestimate exogenous carbohydrate oxidation, but the size of the overestimate cannot be determined from the published data.

How large could the error be?

The letter does not quantify the combined effect. The LJMU group has argued in an earlier paper that when the drink is spiked with highly enriched carbon-13, fluctuations in the background become too small to influence the calculation (5). The argument is correct in principle. But the amount of ¹³C tracer added in some studies was too small to make the effect of background shifts negligible. Even when a very hard exercise session is performed a week before a trial and the diet is controlled in the days leading up to the test with no foods that are derived from C4 plants, there will still be ¹³C in the glycogen stores that will appear as soon as you exercise. Based on the typical background shifts observed over many research projects, we estimate that studies that did not apply a proper background correction overestimate values by about 0.07 g/min or 3-5%. This 3-5% is likely a best-case scenario, because it was calculated from studies in which glycogen stores were depleted 1 week before and the diet was controlled for 7 days, whereas most studies have no glycogen depletion protocol and only a couple of days of diet guidance.

One way to judge the size of the issue is to compare numbers across laboratories. An independent laboratory measured 11 highly trained cyclists at 90 and 120 g/h and corrected the background with extra trials using low-carbon-13 carbohydrate (6). At 120 g/h in a 1:0.8 ratio, exogenous oxidation in the last hour was 1.51 g/min and the efficiency was 76%, against 86% at 90 g/h. An LJMU study in 2022 reported peak oxidation of 1.56 to 1.66 g/min and 72 to 75% efficiency at 120 g/h in nine cyclists (5). These largely agree. The 2025 and 2026 studies are higher. The marathoners reached 1.68 g/min and 84% efficiency, and in the 90, 120 and 150 g/h study efficiency at 120 g/h rose to 89% in the third hour, which is about 1.8 g/min (9,10). In any case when one lab produces consistently higher numbers than studies from other labs (see figure), it should be explained and a possible explanation is the methodology or equipment used. Other explanations could be the subject groups recruited (maybe they were more trained to use carbohydrates during exercise?).

How should we read these papers?

None of this means that the studies are worthless. Claims such as "these are the highest values ever reported" or "we can reach x g/min" need to be questioned. In studies that have not made sure that no assumptions were violated, the outcome is uncertain, and the absolute size of the numbers and the efficiencies need to be interpreted with caution. I hope this blog makes clear that this is not simply measuring heart rate: the procedures need careful consideration and an understanding of tracer methodology to prevent small mistakes or shortcuts that can add up to relatively large errors and sometimes even wrong conclusions.

Practical takeaways

• Do not compare absolute oxidation rates or efficiencies across laboratories. Compare doses within one study.

• Treat the oxidation rates and efficiencies of the newest studies (1.7 to 1.8 g/min, 84 to 89% efficiency) as upper estimates.

• Look at how the background was determined. A trial with water or the ingestion of a carbohydrate with low ¹³C abundance in the same participants is better than a resting breath sample.

• Check whether dose and composition were changed together, because then the effect of dose cannot be isolated.

• Remember that oxidation is a mechanism and not an outcome. The question for athletes is whether performance improves, and that is discussed in the companion blog Is more carbohydrate better? .

References

1. Podlogar T, Cooper-Smith N, Gonzalez JT, Wallis GA. Personalised carbohydrate feeding during exercise based on exogenous glucose oxidation: a proof-of-concept study. Perform Nutr 1(1) 2, 2025. doi:10.1186/s44410-025-00003-9.

2. Pirnay F, Lacroix M, Mosora F, Luyckx A, Lefebvre P. Glucose oxidation during prolonged exercise evaluated with naturally labeled ¹³C glucose. J Appl Physiol 43(2) 258-261, 1977.

3. Jentjens RL, Jeukendrup AE. High rates of exogenous carbohydrate oxidation from a mixture of glucose and fructose ingested during prolonged cycling exercise. Br J Nutr 93(4) 485-492, 2005.

4. Jeukendrup AE, Wallis GA. Measurement of substrate oxidation during exercise by means of gas exchange measurements. Int J Sports Med 26(Suppl 1) S28-S37, 2005.

5. Hearris MA, Pugh JN, Langan-Evans C, Mann SJ, Burke L, Stellingwerff T, Gonzalez JT, Morton JP. ¹³C-glucose-fructose labeling reveals comparable exogenous CHO oxidation during exercise when consuming 120 g/h in fluid, gel, jelly chew, or coingestion. J Appl Physiol 132(6) 1394-1406, 2022.

6. Podlogar T, Bokal S, Cirnski S, Wallis GA. Increased exogenous but unaltered endogenous carbohydrate oxidation with combined fructose-maltodextrin ingested at 120 g/h versus 90 g/h at different ratios. Eur J Appl Physiol 122 2393-2401, 2022.

7. Saris WH, Goodpaster BH, Jeukendrup AE, Brouns F, Halliday D, Wagenmakers AJ. Exogenous carbohydrate oxidation from different carbohydrate sources during exercise. J Appl Physiol 75(5) 2168-2172, 1993.

8. Podlogar T, Rowlands DS. Does 120 g/h carbohydrate ingestion really confer a metabolic advantage in elite marathoners? Methodological concerns and alternative interpretations. J Appl Physiol 140(6) 1802-1803, 2026.

9. Ravikanti S, Silang KG, Martyn HJ, Johnson KO, Louis JB, Bampouras TM, Owens DJ, Jones AM, Morton JP, Pugh JN. ¹³C-labelled glucose-fructose show greater exogenous and whole-body CHO oxidation and lower O₂ cost of running at 120 versus 60 and 90 g/h in elite male marathoners. J Appl Physiol 139(6) 1581-1595, 2025.

10. Martyn HJ, Gonzalez JT, Stellingwerff T, Burke LM, Owens DJ, Pugh JN, Morton JP. Effects of 10, 90, 120 and 150 g/h CHO on substrate metabolism, exogenous oxidation and exercise capacity: implications for CHO personalization. Med Sci Sports Exerc 2026.