Modification of baseline status to improve breath tests performance

Breath tests used to evaluate carbohydrates malabsorption require baseline H2 and CH4 levels as low as possible. Test cancellation is recommended when exceeding certain cut-offs (H2 ≥ 20 ppm and CH4 ≥ 10 ppm). Although following preparation protocols, many patients have baseline levels above those cut-offs. We investigated if light walking can reduce baseline H2 and CH4 levels. We retrospectively analyzed baseline H2 and CH4 levels from 1552 breath tests. Baseline levels (B1), especially in H2, were lower when obtained at later hours of the day. In those with baseline levels above cut-off, re-sampling (B2) after light walking for one hour, decreased H2 levels 8 ppm (Q1–Q3: 1–18 ppm), and 2 ppm (Q1–Q3: 0–3 ppm) for CH4. Consequently, 40% of tests with elevated B1 levels, presented B2 levels below mentioned cut-offs. Ten percent of tests considered negative when using B1 for calculations, turned positive when using B2 instead. All positive tests when using B1 values, remained elevated when using B2. Re-sampling after light walking for one hour could allow test performance in those with previous elevated baseline levels, avoiding diagnosis delays. Using the second sample for delta calculations identifies positive patients for malabsorption that would have been considered negative.

Malabsorption of carbohydrates can provoke several gastrointestinal symptoms including diarrhea, bloating, flatulence and abdominal pain due to intestinal fermentation of these non-absorbed sugars. Multiple pathologies can lead to this malabsorption, such as small intestine bacterial overgrowth (SIBO) 1 , lactase enzyme deficiency causing lactose malabsorption 2 , or the overload of fructose intestinal transporters (GLUT-2 and GLUT-5) causing fructose malabsorption 3 .
Currently, diagnosis of carbohydrates malabsorption is mainly based on breath tests. These null invasiveness tests consist on an oral load with the corresponding sugar and the monitoring of the altered absorption by measuring the fermentation gases (H 2 and CH 4 ) in exhaled air. An increase in H 2 and/or CH 4 levels reflects the intestinal fermentation of non-absorbed sugars. Rise in H 2 or in CH 4 levels depends on the type of anaerobic bacteria present. Although more prevalent intestinal bacteria produce H 2 , some patients present methanogenic bacteria that consume this H 2 to produce CH 4. As a consequence, both H 2 and CH 4 are recommended to be analyzed 4 .
To avoid false negative results, H 2 /CH 4 baseline levels should be low 5,6 and interpretation of elevated baseline levels is still uncertain 7 . We have previously shown that baseline high H 2 levels are associated with more positive lactose breath test 8 . For this reason, patient preparation is required 7,9 , and it includes avoiding drugs such as antibiotic or laxatives, in the previous days, or procedures that could alter colonic microbiota such as colonoscopy. Besides, in the 24 h before the test, patient must follow a diet without fermentable fibers 10 , and the test must be performed in fasting state. In fasting conditions, baseline H 2 levels are usually 7 ± 5 ppm 11 . However, elevated baseline levels are found in a significant number of patients due to different causes, such as SIBO, transgression in patient preparation, or constipation in the case of CH 4 12,13 . For example, Kumar et al. 14 showed that patients with irritable bowel syndrome have a baseline H 2 concentration double than healthy population, while constipation correlate with baseline CH 4 higher than 10 ppm 15 . In fact, some guidelines recommend not to continue the test if baseline levels exceed certain cut-offs ranging 15-20 ppm for H 2 5,7 . This results in a delay of analytical results and diagnosis, and consequently in a detriment of patients. However, a recent survey stated that despite recommendations breath tests are still executed with a wide variety of conditions both in performance and in interpretation 16 . Although the requirement for baseline CH 4 levels as low as possible also applies to CH4, guidelines do not clearly include a specific cut-off value 5,7 .
With the aim to decrease the number of high baseline H 2 and/or CH 4 levels detected during the performance of breath tests, we evaluated the effect of the time of procedure and a previous light walk.

Results
Basal breath test and patients characteristics. Median baseline levels of H 2 were 6 ppm (Q1-Q3: 3-11 ppm) and 4 ppm (Q1-Q3: 2-10 ppm) for CH 4 . Considering the baseline cut-offs for H 2 and CH 4 of 20 and 10 ppm respectively, 11% of tests presented an elevated baseline H 2 level and 25% an elevated baseline CH 4 level. It should be noted that in 2% of tests, both H 2 and CH 4 were above their corresponding cut-offs (Table 1).
Regarding the age of patients, we observed higher baseline H 2 levels in tests performed in underage patients (9 ppm; Q1-Q3: 4-17 ppm) than those in patients over 18 years (5 ppm; Q1-Q3: 3-10 ppm; p < 0.01; Table 1). On the contrary, in the case of CH 4 , levels were higher in adults (5 ppm; Q1-Q3: 2-11 ppm) than in underage patients (4 ppm; Q1-Q3: 2-7 ppm; p < 0.05). These differences were also reflected in the percentage of tests with baseline levels above the cut-off. In the case of H 2 , elevated baseline levels were observed in 22% of tests performed by underage patients, while this percentage decreased to 8% (p < 0.01) in those performed by older patients. Tests with elevated CH 4 baseline levels were 27% of tests in patients above 18 years and only 16% of tests in underage patients (p < 0.01).
When considering the sex of patients, no differences were observed in baseline H 2 levels or in the percentage of tests with baseline levels above the cut-off (Table 1). Also, female patients had slightly higher CH 4 levels (5 ppm, Q1-Q3: 3-11 ppm) than male patients (4 ppm; Q1-Q3: 2-8 ppm; p < 0.01). Accordingly, we observed a higher percentage of tests with elevated baseline CH 4 in female (28%) than in male patients (19%; p < 0.01).
Basal breath test and time of the day. We also investigated the influence of the time of the day in baseline levels. We observed that baseline H 2 levels changed with the hour of testing (Fig. 1a), being significantly higher (p < 0.01) in those performed before 10 am (6 ppm, Q1-Q3: 3-12 ppm) than those performed later (5 ppm, Q1-Q3: 2-9 ppm). Same occurred for CH 4 levels (Fig. 1b), with higher values in tests performed in the early hours (5 ppm, Q1-Q3: 3-10 ppm) than in those performed after 10 am (4 ppm, Q1-Q3: 2-9 ppm).
As mentioned before, 240 patients performed multiple breath tests in different days. The range for baseline H 2 levels in each patient among different days presented a median value of 4 ppm (Q1-Q3: 2-8 ppm), higher than 2 ppm observed for CH 4 (Q1-Q3: 1-4 ppm; p < 0.01 Fig. 2a). The difference observed in H 2 baseline levels did not depend on the hour of the day the sample was taken in or how distance were the different sampling hours (Fig. 2b). Median coefficient of variation of baseline levels among patients was 57% (Q1-Q3: 28-88%) for H 2 and 35% (Q1-Q3: 16-71%) in the case of CH 4 . Table 1. Baseline H 2 and CH 4 levels according to age and sex of patients, indicated as median and interquartile range (Q1-Q3) is shown in a bracket. Baseline samples classification according to H 2 and CH 4 levels compared to their corresponding cut-offs. H 2 high means ≥ 20 ppm, CH 4 high means ≥ 10 ppm, H 2 low means < 20 ppm and CH 4 low means < 10 ppm. Comparing with patients above 18 years, * represents p < 0.01 and ~ represents p < 0.05. # represents p < 0.01 compared with female patients. www.nature.com/scientificreports/ This variation observed in baseline H 2 levels obtained in different days, resulted in that 16% of patients presented an H 2 baseline level below cut-off in one test and above that cut-off in another test performed in a different day (p < 0.01).
Breath test performance after baseline repetition. In 126 test with initial baseline (B1) levels above cut-off, after light walking and re-sampling (B2), tests continued with stimulus intake. If first baseline B1 was considered for delta calculations, H 2 breath test was negative in 73% of the tests. However, 10% of these turned positive when delta was calculated with the repeated baseline B2 (Table 3). On the contrary, all the tests positive using B1, remained elevated when using B2 for calculations. This differences in classification reached statistical significance (p < 0.01).
Related to CH 4 , there was an increase in 8 tests when considered the B2 but no increase when using B1 to calculate the delta. On the contrary, in 6 tests with increases related to B1, the increase disappeared when considering B2 for calculations. In this case, differences in classification were also statistical significant (p < 0.01).

Discussion
To our knowledge, this is one of the first studies with larger sample size focused in baseline breath measurements. First of all, we observed the presence of methanogenic microbiota determined by the presence of CH 4 in breath, in a high percentage of the patients, higher than the prevalence previously indicated 17,18 . Levitt et al. 18 found that only 36% of patients presented CH 4 levels higher than 1 ppm. Probably, this difference could be due to the different type of population analyzed. Prevalence of elevated CH 4 , baseline levels in our work was similar to that observed in a study of Harvie et al. 17 who estimated a prevalence of 26% of high CH 4 producers. However, they established 5 ppm as baseline cut-off. If using that value, the percentage of tests with elevated baseline CH 4 in our population levels would have risen from 25 to 49%. This reinforces the indication of measuring both H 2 and CH 4 to avoid false negative results.
Our baseline H 2 is similar to that reported previously 11 . Even when patients follow the indicated preparation, we have observed that baseline levels often exceed recommended cut-offs. When analyzing their demographic data, we found that those baseline levels were different between adult and underage patients. Since all samples included in the study had adequate CO 2 levels, differences in H 2 and CH 4 levels cannot be attributed to an incorrect breath sampling technique in pediatric patients. Besides, H 2 and CH 4 performances are indeed opposite, which rules out contamination as the reason of the observed differences. On one hand, H 2 baseline levels were higher in minor patients with a higher percentage of patients with baseline levels above the recommend cut-off. Le Neve et al. 19 did not observe this association with patients' age although their study was performed only in Table 3. Performance of breath tests with baseline re-sampling. Tests were classified according to the presence of an increase in H 2 or CH 4 levels after stimulus intake, depending on whether first baseline of repeated baseline were considered for delta calculations. Increase in H 2 levels was considered when ≥20 ppm and in CH 4 levels when ≥10 ppm.   20 , which has been related to CH 4 production, and other intestinal alterations such as diverticulosis 21 . We can reject the confounding effect of smoking in the results as patients were instructed not to smoke in the day of test performance and previous smoking effect in H 2 and CH 4 levels disappears after 10-15 min 22,23 .
Concerning time of day, we observed that H 2 levels fluctuated more than CH 4 levels, with higher levels earlier in the morning. To check if this was just a consequence of interindividual variability we focused in those patients with multiple tests in different days. We found that CH 4 was much more stable than H 2 , contrary to that described by Jonderko et al. 24 . This could be due to sample size differences since our study comprised 240 patients compared to only 12 of Jonderko et al. In any case, H 2 range across different days was reduced (although not significantly) when looking in those tests performed at same time of day (less than 60 min apart), suggesting that at least some of the variability observed can be attributed to the time of the day the test is performed. Shibata et al. 25 found in healthy controls slightly higher levels (an increase of 2 ppm) of fasting H 2 in the morning than before sleep the previous night. Related to this, we should consider that during the night there is a nocturnal hypoventilation that can cause these slightly elevated fasting levels of breath H 2 produced by persisting fermentable substrates in the colon 26 .
Vigorous exercise is not recommended prior breath tests because it can alter ventilation rate and subsequently exhaled H 2 and CH 4 and its relation to CO 2 27 . However, our indication is to go for a walk, light enough to not alter ventilation rate before and during the test. Even more, there was a minimum 5 min delay between patient returns and repeated sampling, and a visual inspection of potential hyperventilation. This light walk resulted in a reduction of baseline levels (mainly in H 2 levels) that allowed to perform 40% of these tests. Interestingly, all tests in which only CH 4 baseline levels were elevated, CH 4 continued also elevated in the repeated baseline sample. Consequently, it seems unnecessary to perform the light walk and resampling when only CH 4 is elevated in baseline samples.
The decrease in H 2 levels after the light walk and the influence of time of day may be indeed related, as light activity could induce intestinal motility and feces stirring causing the release of preformed H 2 trapped in the feces 28,29 . After one hour, this H 2 would have been already released and the baseline levels would reflect more precisely the baseline state.
To date, it is not yet clear if elevated H 2 baseline levels are related with diet transgressions, reflect SIBO presence 30,31 , or even pancreatic alterations such as exocrine insufficiency 32 or pancreatic duct stenosis 33 . Although our protocol includes oral washing to avoid oral microbiota interference and patients are also questioned about potential diet transgressions that could affect breath performance, we have found that elevated baseline H 2 levels were quite frequent (11%). However, they are critical since (i), when occurred, test cancelation is recommended and diagnosis is delayed and (ii), in the case of continuing, they can conceal a higher increase in H 2 levels than those reflected in delta calculations, and thus lead to false negative results. In fact, we have showed that in those tests with high baseline levels and negative result, 10% of them would turn positive if the delta value were calculated with the repeated baseline level after light walking instead. However, all tests with high baseline levels and positive result would remained positive when repeated baseline was used to calculate the delta value.
In summary, breath status can be modified in patients by light walking, which reduces H 2 and CH 4 baseline levels. This allows tests continuation and avoids tests re-scheduling and diagnosis delays in a significant percentage of patients.

Methods
Patients. We retrospectively analyzed 1552 breath tests, from 1305 patients (Table 4), 756 (49%) fructose breath test and 796 (51%) lactose breath test. Of these, 232 (18%) performed two breath tests in different days and 8 (0.5%) performed three breath tests. Those tests were performed with a median of 9 days apart (Q1-Q3: Breath tests performance. Patients were required to perform the test in fasting and without having smoked that day. They had to accomplish a diet the previous day without fermentable fibers, lactose or fructose. Adherence to prescribed diet was checked before the beginning of the test. Fructose and lactose breath tests were not performed if patients had recently received antibiotics or laxatives or procedures such as colonoscopy. Prior to breath sampling, adults patients rinsed their mouths with a 10 mL of commercial oral antiseptic solution containing chlorhexidine. In underage patients rinses were performed with water. The stimulus were 25 g of fructose or lactose, respectively. In the case of pediatric patients, dose was adjusted to 1 g/kg with 25 g as the maximum dose. Breath baseline samples and each 30 min for three hours after stimulus intake were obtained and 20 mL of them were analyzed for H 2 and CH 4 levels (ppm) in a Breath Tracker SC (Quintron, Milwaukee, USA). CO 2 levels (in %) were also assayed simultaneously in the same analyzer to evaluate potential contamination of samples with ambient air 34 . If any sample presented CO 2 < 2%, sample results were rejected and a new sample was obtained. The breath analyzer was daily calibrated and verified before processing any sample, using the gas calibrator provided by manufacturer. Our data indicate that imprecision and bias were respectively 3.1% and 4.7% for H 2 measurement, and 3.4% and 4.1% for CH 4 .
According to guidelines, breath test lacks of reliability if baseline H 2 exceed 20 ppm. If baseline samples (B1) rendered H 2 above 20 ppm, a new baseline sample should be obtained. According to our own protocol and in order to achieve levels below that threshold, patients were indicated to go for a light walk outdoors for one hour with no goal in terms of distance. Although no specific cut-off is established for CH 4 in guidelines, our protocol indicates the same procedure if baseline CH 4 exceeds 10 ppm. The walk should be light enough to not provoke any stress, fatigue or hyperventilation. After arrival there were about 5 min of resting and a visual inspection of potential hyperventilation, before a new baseline sample (B2) was obtained. In the case that levels kept above the cut-off, breath test was cancelled or re-scheduled. Tests were considered positive when H 2 or CH 4 levels increased above baseline: 20 ppm in the case of H 2 and 10 ppm in the case of CH 4 . Data analysis. Data were statistical analyzed with SPSS v20 software (IBM). Data are indicated as median an interquartile range (Q1-Q3). According to the normality tests performed (Shapiro-Wilk and Kolmogorov-Smirnov), both initial and repeated H 2 and CH 4 baseline levels followed non-normal distributions. Therefore, comparison between different patients were performed with Mann-Whitney U test whereas comparison between samples of the same patient was performed with Wilcoxon signed rank test. Frequencies were compared with χ 2 test. P < 0.05 was considered as statistically significant.