Article Text


European interlaboratory comparison of breath13CO2 analysis


The BIOMED I programme Stable Isotopes in Gastroenterology and Nutrition (SIGN) has focused upon evaluation and standardisation of stable isotope breath tests using 13C labelled substrates. The programme dealt with comparison of 13C substrates, test meals, test conditions, analysis techniques, and calculation procedures. Analytical techniques applied for13CO2 analysis were evaluated by taking an inventory of instrumentation, calibration protocols, and analysis procedures. Two ring tests were initiated measuring 13C abundances of carbonate materials.

Evaluating the data it was found that seven different models of isotope ratio mass spectrometers (IRMS) were used by the participants applying both the dual inlet system and the continuous flow configuration. Eight different brands of certified 13C reference materials were used with a 13C abundance varying from δ13CPDB −37.2 to +2.0 ‰. CO2 was liberated from certified material by three techniques and different working standards were used varying from −47.4 to +0.4 ‰ in their δ13CPDB value.

The standard deviations (SDs) found for all measurements by all participants were 0.25 ‰ and 0.50 ‰ for two carbonates used in the ring tests. The individual variation for the single participants varied from 0.02 ‰ (dual inlet system) to 0.14 ‰ (continuous flow system). The measurement of the difference between two carbonates showed a SD of 0.33 ‰ calculated for all participants. Internal precision of IRMS as indicated by the specifications of the different instrument suppliers is <0.05 ‰ for dual inlet systems and <0.3‰ for continuous flow systems. In this respect it can be concluded that all participants are working well within the instrument specifications even including sample preparation. Increased overall interlaboratory variation is therefore likely to be due to non-instrumental conditions. It is possible that consistent differences in sample handling leading to isotope fractionation are the causes for interlaboratory variation. Breath analysis does not require sample preparation. As such, interlaboratory variation will be less than observed for the carbonate samples and within the range indicated as internal precision for continuous flow instruments. From this it is concluded that pure analytical interlaboratory variation is acceptable despite the many differences in instrumentation and analytical protocols. Coordinated metabolic studies appear possible, in which different European laboratories perform13CO2 analysis. Evaluation of compatibility of the analytical systems remains advisable, however.

Statistics from


13CO2 breath tests are attractive due to the use of non-radioactive substrates and the simple non-invasive sampling technique. A 13CO2 breath test is performed by administration of a 13C labelled substrate which is finally metabolised to 13CO2. The rate of 13C excretion in breath is determined by the rate limiting step in the overall process of use. The rate limiting step is located at the site of an impaired organ or enzyme function. In this way 13CO2 breath tests are principally ideal tools for diagnosis of digestive or metabolic dysfunction. The biological function to be tested can be of various origins, such as:

  • Organ function (gastric emptying rate, liver function, pancreas function, intestinal integrity)

  • Digestion, absorption, or metabolism of nutrients (carbohydrates, fat, protein)

  • Bacterial colonisation (stomach, small intestine).

 The kinetics of 13C exhalation in breath is measured in consecutive breath samples collected during the experimental period. On most occasions the 13C recovery in breath is mathematically related to the dose given, leading to a quantitative description of the processes involved in substrate use. In certain other cases the time course of 13C exhalation is described in terms of lag phase, t1/2, and appearance rate (test for gastric emptying rate) or the 13C enrichment itself is considered the final diagnostic value (urea breath test for detection of gastric Helicobacter pylori infection).

Many steps, however, are involved in total use of substrate. All these steps are subjected to intraindividual and interindividual variation and site of rate limiting step may vary from one healthy individual to another. This changes the apparently simple breath test into a complex system in terms of interpretation. For application in nutritional and clinical research and clinical diagnostics, the rate limiting steps in digestion, absorption, and metabolism of various substrates have to be analysed. At the non-physiological level the type and dose of13C substrate and analytical aspects of measuring13C abundances in breath carbon dioxide have a dominant role.

Substrate dose: physiological and analytical considerations

The substrate dose needs definition in terms of total mass and of13C mass. Breath tests meant to monitor digestive or absorptive capacity require a physiological load of nutrient. On the other hand, metabolic functions can generally be tested by applying a tracer dose of substrate. The 13C content ingested is determined by the expected degree of digestion, absorption (oral tests), and metabolism as well as the sensitivity of the analytical technique used. The origin of 13C labelled compounds can be (bio)chemical synthesis (selective high degree of labelling, >95% at one or more carbon positions) or natural (random low degree of labelling, 0.02% per average C-atom). Naturally, 13C labelled material is produced by C4 plants. Owing to different photosynthetic pathways the carbon isotope fractionation in these plants is less than in C3 plants resulting in a slightly higher 13C abundance (1.10% v1.08%). Compounds isolated from C4 plants can be used in metabolic studies in Europe where diets are mainly based on C3 plants. In the case of a loading test using highly enriched 13C substrate, this substrate is “diluted” with unlabelled substrate to a product with the correct amount of mass and the correct amount of 13C. Naturally, 13C enriched substrates are used without dilution and only for loading purposes. The total amount must suit the need for mass and13C. The 13C content of the substrate must be sufficient to create a detectable increase in 13C abundance in breath CO2. For this purpose, 13C abundance in breath should exceed more than two to three times the SD of the basal value before the experiment.

Measurement of 13CO2 enrichment in exhaled breath

13C is a naturally occurring isotope consisting of approximately 1.1% of total carbon. The 13C isotope is not radioactive and can be detected by a number of techniques such as mass spectrometry, nuclear magnetic resonance spectroscopy (NMRS), infrared spectroscopy, and laser resonance spectroscopy. For13CO2 analysis, mass spectrometry is the traditional technique, whereas infrared and laser resonance spectroscopy are developing alternative techniques. With regard to mass spectrometry, special equipment is required to measure13CO2 enrichment owing to the high accuracy needed for low level enrichment (0.001–0.01 at %) that is generally obtained in breath tests. The required accuracy is obtained with IRMS originally developed for geological applications measuring variation in natural 13C abundance. 13C abundance in nature is not constant but dependent on the origin of the carbon source varying from around 1.06% for fossil fuels to 1.12% for carbonate stones. The natural background of 13C in breath CO2 is 1.08–1.09% depending on the composition of C3 and C4 plant material in the diet. In Europe the 13C abundance is between 1.080 and 1.085% (Nakamura K, Schoeller DA, Winkler FJ, Schimdt HL. Biomed Mass Spectrom1982;9:390) The enrichment in 13C during a breath test depends on the dose and 13C enrichment of the substrate and the degree of metabolic 13CO2production. Generally, the dose given results in a maximum absolute increase in 13C abundance of breath CO2 of about 0.005–0.05%. To enable suitable precision, stability, and comparable data, 13C abundance is always measured against a universal reference standard, this being the carbon originating from Pee Dee Belemnite Limestone (PDB). Data are expressed as per mil (‰) difference to PDB: δ13CPDB (‰).13C enrichment is defined as the difference between the basal 13C abundance before administration of substrate and the 13C abundance at a certain time point after administration and is expressed as Δδ13CPDB(‰). Biological variation of baseline δ13CPDB values is in the order of SD 0.5 (‰) .


The accuracy of the 13C abundance measurement is related to a number of variables.


As mentioned, IRMS, infrared, or laser resonance techniques can be used. Two types of IRMS equipment can be distinguished—namely, dual inlet systems in which each sample CO2 and reference CO2 are measured in sequence up to 10 times in a row and continuous flow systems in which sample CO2 and reference CO2 are measured singly in one or in separate runs.

Inlet system

Breath gas can be introduced directly as is done with infrared and laser resonance techniques, or after separation of CO2 from the other breath gas components as is done for IRMS analysis. CO2 can be isolated from breath by cryogenic trapping (on line or off line) or by gas chromatography. The first technique is used in combination with a dual inlet IRMS, the gas chromatography technique with continuous flow IRMS.

Quality of reference compounds

Isotope ratios are measured by IRMS against a working CO2 standard which is calibrated against a certified reference standard, which in turn is calibrated against PDB limestone. Different working reference standard CO2 sources can be used as well as different certified references with known13C abundances.

Calibration procedures

Quality of measurements is affected by the accuracy of instrument calibration.

Aim of interlaboratory comparison

The general aim of the BIOMED I programme was the comparison and possible coordination of breath test procedures. This was done at the level of choice of type and dose of substrate, experimental conditions, definition of end variables, and the level of 13C enrichment measurements. For the level of 13C enrichment measurements, the types of instruments, calibration procedures, and analysis procedures have been compared. Applying circulating test compounds actual 13C abundance measurements were compared.

Plan of investigation


A questionnaire was sent to the different laboratories involved in the BIOMED I programme and equipped with IRMS or alternative techniques to measure 13C enrichment of CO2. A total of 15 laboratories participated in the survey and one or both of two ring tests. Only IRMS instruments were used. The questions asked dealt with types of instruments available at site, instruments used for the ring test measurements, types of reference compounds, and calibration procedures as well as the analytical procedures for sample analysis.


A ring test was set up. This was organised by a collaborative action of the Laboratory for Nutrition and Metabolism at Groningen University Hospital, Groningen, the Netherlands (Dr F Stellaard, Prof Dr RJ Vonk) and the Laboratory for Digestion and Absorption of the University Hospital Gasthuisberg, Leuven, Belgium (Dr B Geypens, Prof Dr Y Ghoos). Participants received a code denoted by Prof Dr W van der Slik (University Hospital Groningen, Department of Clinical Chemistry). Results were delivered to the organisers under strict code. Test compounds selected by the organisers were shipped from Leuven, Belgium. For the purpose of stability, test compounds selected were (bi)carbonates with different 13C abundances. The samples were delivered to the participants together with solid phosphoric acid and tubes and a detailed instruction sheet describing a standard protocol of how to liberate CO2 from the (bi)carbonate material. The 13C abundance of the CO2 formed could be determined with the regular protocol used in the participants’ laboratory. Participants were asked to do triplicate analyses of three aliquots of the sample. The mean and SDs of nine values were compared. 13C abundance is presented as the δ13CPDB (‰) data.


At first a carbonate sample and a bicarbonate sample were distributed. However, the release of CO2 from the bicarbonate sample appeared difficult to control leading to variable gas pressures and possible isotope fractionation. Owing to large standard errors (SEs) the bicarbonate data are omitted from the results section.


Secondly, a ring test was done applying two carbonate samples with different 13C abundances (codes BG02 K and BG03 N). For comparison, the individual 13C measurements and the difference between the two mean values were compared. The difference was considered more valid because it reflects the real situation where baseline 13C abundance is subtracted from the enriched value.


Table 1 shows the laboratories that participated in the survey of instrumental information and the two ring tests. Nine laboratories participated in ring test 1 and nine in ring test 2.

Table 1

Participants of the European interlaboratory comparison of breath 13CO2 analysis



All instruments used in the participating laboratories were manufactured by one of three suppliers: Finnigan MAT, Bremen, Germany; Micromass, Manchester, UK; and Europa Scientific, Crewe, UK. No instruments other than mass spectrometers were used. A total of 11 different models were listed as being available (table 2). Of these, seven different models were applied during the ring tests. On three occasions the instrument used for ring test purposes was equipped with a dual inlet system, providing the highest accuracy and reproducibility (internal variation <0.005%). In all other cases a continuous flow system was applied (internal variation <0.3%). For daily use a working standard is used as reference material. This consists of a gas cylinder with CO2 (n=6) or CO2 in helium (n=1). The δ13CPDB value varied from −47.41 to +0.44‰. One participant used breath as a working reference. Eight different brands of certified reference standards are being used by the participants (table 3). The δ13CPDB values varied from −37.22 to + 1.87‰.

Table 2

IRMS instruments

Table 3

Calibration standards used for breath13CO2 analysis

Analytical procedures

Breath collections were done in vacutainers or exetainers (10–15 ml) with the exception of one participant who used collection bags (TECO, Alu). The laboratories applying dual inlet IRMS instrumentation prepared the breath samples by on line and off line cryogenic procedures introducing clean CO2 through the sample port of the dual inlet system. The laboratories applying continuous flow IRMS instruments used three different types of gas chromatographic separation columns and at least as many temperature programmes. The working reference gas was analysed after a set of samples (1–10). In all but one laboratory, drift corrected data were used for further calculations. Calibration of the working standard is done from once a week (breath) to two to three times each year. On most occasions calibration is done after renewal of the CO2 gas cylinder. Certified solid reference material is prepared by acid liberation of CO2 or through combustion by off line or on line techniques.

Ring test 1

The mean δ13CPDB of 81 individual measurements obtained for the carbonate sample was −11.32 ‰. The SD was 0.33 ‰. The mean of nine measured values for the individual laboratories varied from −12.28 to −10.30 ‰. The SD varied from 0.02 to 0.14 ‰. Two laboratories produced mean values outside the mean (SD 2) range. The results of this first ring test were presented to the participants under code during a BIOMED 1 meeting. The outliers were stimulated to check their analytical protocol. The results also led to the idea that measurement of the difference between two test compounds is more representative of the real situation. A system with a general offset of measured δ13CPDB values may still give good values with respect to enrichment values.

Ring test 2

The means (SD) of δ13CPDB values for the nine participating laboratories were −19.48 (0.38) ‰ and −5.61 (0.63) ‰ for both carbonate samples (fig 1 and 2a). In both cases, two participants were outside the mean (SD 2) range. The difference was 13.87 (0.41) ‰ (fig 2b). With respect to the difference only one participant was outside the mean (SD 2) range.

Figure 1

Overall test results of 13C abundance measurements in CO2 liberated from two carbonate sources (BG03 N and BG02 K) (n=9).

Figure 2

Interlaboratory comparison of 13C abundance measurements of CO2 liberated from two carbonate sources (BG03 N and BG02 K): (A) individual δ13CPDB values and (B) difference between the values for carbonates 1 and 2.


The survey revealed that a large number of analytical variables were present in the instrumentation (suppliers, models, inlet system, procedures) and calibration procedures (working standard, reference materials) used by the participants in the ring tests. This is expected to induce a large variation in mean values obtained by the different participants. The SDs of the measured values were 0.33 for ring test 1 and 0.38 and 0.63 ‰ for the two individual carbonate test compounds in ring test 2. The variation within one laboratory (0.02–0.14 ‰) was much lower than the interindividual laboratory variation (ring test 1, 0.33 ‰). To judge these values, they should be put into perspective by comparing them with pure instrumental variation. Internal variation for 13C abundance measurements specified for IRMS instrumentation by the different mass spectrometry companies is normally <0.05 per mil for dual inlet instruments and <0.3 per mil for continuous flow instruments. Furthermore, preparation of CO2 from the carbonate samples also introduces some degree of variation due to isotope fractionation. Total variation is less than twice the instrumental variation of continuous flow instrumentation, whereas individual variation was within the specification criteria of continuous flow instrumentation. The difference between the two values obtained for both carbonates showed intermediate variation inbetween both single measurements. The result of only one participant was outside the mean (SD 2) range. It is a question whether the outlier results from instrumental deviations or a deviation due to sample preparation. This question might be solved by another ring test measuring 13C enrichment of actual CO2 gas at a physiological breath concentration (3–4%) in a N2/O2 mixture (80/20 v/v). Stability of samples needs to be tested, however, and different sample sizes need to be offered to fit the needs of the individual instruments.


The variation of the 13C enrichment values obtained by the participating laboratories obtained with CO2 liberated from carbonates is within the specifications of continuous flow instrumentation. Actual instrumental and procedural variation may be even smaller. It appears feasible to have more laboratories do breath13C analysis in a coordinated European metabolic study. However, compatibility of the analytical systems should be checked beforehand. Standardisation of analytical protocols appears to be unnecessary, provided that each individual laboratory ensures regular calibration of the system and an internally standardised analytical procedure.

View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.