Introduction

Delayed gastric emptying (GE) and intolerance to enteral feeding occurs in over 50 % of critically ill patients [1, 2]. Not only does this result in inadequate nutritional support, but it is also a risk factor for gastro-oesophageal reflux and aspiration [3–5]. The ability to detect delayed GE early in these patients may help clinicians to institute treatment promptly to minimize associated complications [3]. The presence of gut dysfunction during critical illness, such as delayed GE and feed intolerance, is associated with poorer outcomes and has been proposed as a surrogate marker of a more severe illness [2–5]. Gastric emptying is also measured for research purposes, for example, to quantify the effects of putative gastric prokinetic agents with the rationale of improving the delivery of naso-gastric nutrients.

A number of techniques have been used to measure gastric emptying in critically ill patients and all of these have limitations. Although scintigraphy is the gold-standard, its application has been restricted for a number of reasons including radiation exposure, interference with patient care, and technical demands that include the need to transport the patient out of the intensive care unit (unless a portable gamma camera is available). For these reasons, several indirect methods for assessment of gastric emptying have been used. Monitoring of gastric residual volume (GRV) is the most common method and has been used clinically for many decades to guide the provision of enteral nutrition [6]. However, the data supporting its relationship to either gastric emptying or pulmonary aspiration are limited and controversial [6, 7]. The paracetamol absorption test is another indirect technique for assessing gastric emptying and is based on the assumption that the rate of paracetamol absorption is directly correlated to the rate of gastric emptying [8]. This technique involves blood sampling at regular intervals and has not been validated against scintigraphy in critically ill patients.

The octanoate breath test was developed as a non-invasive technique to quantify gastric emptying without the need for blood sampling [9]. It has been validated against scintigraphy in healthy subjects for both solid and non-nutrient liquid emptying [9–17] and in a number of disease states including; diabetes, dyspepsia, respiratory disease and gastroparesis [11, 12, 14, 17–21]. In all these studies there was a strong correlation between the results of the breath tests and scintigraphy (r values 0.66–0.97). Low inter- and intra-individual variability of the breath tests results have been reported [22–24]. In addition, when compared to scintigraphy, both the specificity (73–94 %) and sensitivity (67–100 %) of the breath test were acceptable [12, 14, 18, 20, 21]. However, few studies have included subjects with marked gastroparesis, so the performance of the test at very slow rates of GE is unknown. In critically ill patients, we have recently reported a good correlation between the gastric emptying coefficient (GEC) calculated from the 14C-radioactive labelled octanoate breath test technique and gastric retention based on scintigraphic measurement of a liquid nutrient meal [25]. Although the physiological bases of gastric emptying assessment by either 14C- or 13C-octanoate breath test (13C-OBT) are the same, 13C-OBT does not involve radiation exposure and is much more commonly used than 14C-OBT for assessing gastric emptying in the critically ill. Furthermore, the methods of sample collection and analysis for 14C-OBT are different from those of 13C-OBT [1, 9, 25]. In contrast to the 13C-OBT, the 14C-OBT requires the addition of a special solution to the expired breath sample to ensure adequate expiratory CO2 levels [1, 25]. Whilst measurement of 14C is performed with a liquid scintillation counter, 13C is measured with mass spectrometry [1, 25]. Although Ghoos et al. [9] suggested that 14C-OBT and 13C-OBT are interchangeable in healthy subjects, the results have not been replicated elsewhere and thus far, 13C-OBT has not been formally validated against scintigraphy in critical illness. The aim of the current study was to evaluate the accuracy of the 13C-OBT against scintigraphy for the measurement of gastric emptying in critically ill patients.

Methods

Subjects

Concurrent assessments of gastric emptying by gastric scintigraphy and 13C-OBT were performed prospectively in 33 critically ill patients who were admitted to a mixed medical/surgical, tertiary referral intensive care unit (Table 1). Inclusion criteria were for any patient aged at least 17 years, receiving mechanical ventilation and able to receive enteral nutrition. Exclusion criteria included contra-indication to the passage of an enteral tube, a history of gastric, oesophageal or intestinal surgery, recent major abdominal surgery, and administration of prokinetic therapy (including metoclopramide, erythromycin and cisapride) within 24 h prior to study. The study was approved by the Human Research Ethics Committee of the Royal Adelaide Hospital. This study was conducted in accordance with the National Health and Medical Research Council of Australia National Statement on Ethical Conduct in Human Research 2007 and the guidelines for the conduct of research involving unconscious persons. Written, informed consent was obtained from the next of kin prior to study.

Table 1 Demographic characteristics of critically ill patients (data are expressed as mean ± SE)
Full size table

Study protocol

All studies were performed in the morning, after a minimum 6 h fast. Subjects were positioned 30° head elevated and slightly tilted to the right. After verifying the correct position of the naso-gastric (NG) tube [14-Fr Flexiflo (Ross Laboratories, Columbus, OH) or 16-Fr Levin tube (Pharma-Plast, Lynge, Denmark)] by abdominal X-ray, gastric contents were aspirated and discarded. A test meal consisting of 100 ml of Ensure® (Abbott Australasia Pty Ltd; containing 106 kcal with 21 % fat) mixed with 20 MBq 99mTc-sulphur colloid and 100 mcg 13C-octanoate, was instilled into the stomach via the NG tube, over 5 min in order to prevent the forced injection of feed into the duodenum. At the end of the meal infusion, concurrent measurement of gastric emptying by scintigraphy and 13C-OBT were commenced. Scintigraphic data and breath samples were collected over 240 min. On completion of the study, NG feeding was re-commenced as clinically indicated.

Techniques and data analysis

Gastric scintigraphy

The scintigraphic measurement of gastric emptying involved a dynamic study performed in the ICU using a mobile gamma camera (Starcam 3200iXCT, General Electric, US). Patients were studied supine at 30° head elevation and the camera was placed in the left anterior oblique position. Scintigraphic images were acquired for 240 min (3 min frames).

Data were corrected for radionuclide decay and γ-ray attenuation. Regions of interest were drawn around the total stomach, which was subsequently divided into proximal and distal gastric regions. Gastric emptying curves (expressed as percentage retention over time) were derived. The amounts of the meal remaining in the total, proximal and distal stomach at 0, 5, 15, 30, 45, 60, 90, 120, 150, 180, 210 and 240 min were calculated [26]. The scintigraphic gastric half emptying time (t 1/2) could not be determined in approximately one-third of the patients because 50 % emptying was not achieved by 240 min. In the current study, delayed gastric emptying was defined by >13 % meal retention at 180 min [25].

13C-Octanoate breath test

Breath samples were taken prior to test meal administration, every 5 min for 1 h and every 15 min for a further 3 h. End-expiratory breath samples were collected from the ventilator limb using a T-adapter (Datex-Engstrom, Helsinki, Finland) and holder for vacutainers (Blood needle holder, Reko, Lisarow, Australia), containing a needle (VenoJect®, Terumo Corporation, Tokyo, Japan). Previous studies indicate that equilibration of the CO2 concentration between the ventilator limb and evacuated 10 mL tubes (Exetainer®, Buckinghamshire, England) take only a fraction of a second and that it is a reliable technique for breath sampling [27]. To avoid sampling other than end-expiratory air, samples were timed to the end-expiratory phase by observation of the patient and the time-flow curve on the ventilator monitor.

The concentration of CO2 and the percentage of 13CO2 were measured in each sample using an isotope ratio mass spectrometer (Europa Scientific, ABCA model 20\20, Crewe, UK). Samples containing <1 % CO2 were not regarded as representative of end-expiratory gas and were excluded from further analysis. The 13CO2 concentration over time was plotted and the mathematical formulae used to calculate the best fit curves for % dose per hour and % cumulative dose [9, 27] were:

  • Curve fitting to derive % dose/h: \( Y\, = \,(a\, \times \,(x^{b} ))\exp ( - c\, \times \,x) \)

  • Curve fitting to derive % cumulative dose: \( Y = m(1\, - \,\exp ( - k\, \times \,x))^{b} \) From the resultant curves, GEC and gastric half emptying time [t50(BT)] were derived using the following formulae [9, 27]:

  • \( T_{1/2} = [ - 1/k] \, \times \,\ln [1\, - \,2\, - \,1/b] \)

  • GEC = ln(a)

GEC is a global index of the gastric emptying rate, which accounts for both the rate of appearance and disappearance of the label in the breath [27].

Feeding protocol and definition of feed intolerance

For patients who received enteral nutrition support, the caloric requirements and rate of naso-gastric feeding were assessed and prescribed by a dedicated ICU dietician. Feeding was then managed according to a standardized enteral feeding protocol, in which the feeding rate was increased by 20 ml/h every 6 h up to the patient’s prescribed rate (60–100 ml/h) providing 6-hourly GRVs were less than 250 ml. In patients with a GRV <250 ml, aspirated contents were returned via the NG tube. For aspirate volumes ≥250 ml, defined as “feed intolerance”, contents above 250 ml were discarded and the rate of feeding was reduced by 50 % or to a minimum of 20 ml/h. If subsequent GRVs were <250 ml, aspirated contents were returned via the NG tube and the feeding rate increased by 20 ml/h every 6 h, up to the patient’s prescribed rate. The presence of feed intolerance (within the 72 h of the gastric emptying study) was documented.

Statistical analysis

Most data are expressed as mean [standard deviation (SD)]. Data for t50(BT) are expressed as median [inter-quartile range (IQR)]. The relationships and its strength between scintigraphic and breath test measurements were examined by correlation and linear regression analyses. Gastric scintigraphy was considered as the “gold-standard” on the receiver operating characteristic (ROC) curve analysis, with delayed gastric emptying defined as >13 % meal retention at 180 min [25]. The ROC analysis with the area under the curve (AUC) calculation was performed to determine the best cut-off value of breath test parameters to define delayed gastric emptying. The ROC analysis calculates the sensitivity and 1-specificity of each possible cut-off in breath test parameter compared to the gold standard. An optimal cut-off is one that maximises the sensitivity while minimising 1-specificity. As surrounding cut-offs may improve one of these while worsening the other, this is a subjective process informed by the relative costs of false positives and false negatives. This process is limited by a lack of power in the determination of a normal range for gastric emptying using these techniques. Published normal ranges cannot be used as gastric emptying varies depending on many factors including the energy content and macronutrient makeup of the meal, the position, age and gender of the subjects. Based on a power calculation (with 80 % power, α = 0.05, two tail), a sample size of at least 150 subjects would be required to identify a normal range and allow categorical definition of delayed gastric emptying from the breath test. This has not been performed to date. Hence, the determination of a cut-off GEC to define delayed gastric emptying needs to be interpreted with caution. A P < 0.05 was considered as statistically significant in all analyses.

Results

Demographics and characteristics of the patient cohort are summarized in Table 1. All patients tolerated the study procedures well, and no adverse events were recorded. Nineteen patients (58 %) received NG feeds prior to the gastric emptying study. After the study, all patients were given naso-gastric feeding as per protocol. Overall, 23 patients (70 %) had feed intolerance within the 72 h of the gastric emptying assessment.

Based on scintigraphic assessment, delayed gastric emptying was present in 27 (81 %) patients. The mean gastric retention was 45 (5) % at 180 min and 38 (5) % at 240 min. Gastric retention over 240 min correlated positively with age (r = 0.51; P = 0.02) and an APACHE II score on admission (r = 0.40; P = 0.04), but not with gender, BMI, or APACHE II scores on the day of study. There was a trend for greater gastric retention over 240 min in patients who received morphine/midazolam (44 (6) vs. 23 (8) %; P = 0.05), and inotropes (48 (9) vs. 29 (6) %; P = 0.07).

Correlation between 13C-OBT and scintigraphic parameters

Scintigraphic meal retention over 240 min correlated significantly with both GEC (r = −0.63 to −0.74; P < 0.0001) and t50(BT) (r = 0.55–0.66; P < 0.001) (Fig. 1). The highest correlations were observed within the first 120 min (GEC: r = −0.74; P < 0.0001 and t50(BT): r = 0.66; P < 0.001), and correlations were stronger for GEC than t50.

Fig. 1
figure 1

Correlation between meal retention (%) over 240 min and breath test measurements: a gastric half emptying time (min) and b gastric emptying coefficient

Full size image

There was a large variation in t50(BT) measurements [mean (SD) t50(BT) = 127 (108) min]. Five patients had t50(BT) between 400 and 600 min. Given the variability in the t50(BT) and the weaker relationship with scintigraphic meal retention, the t50(BT) was judged to be less useful than GEC for measuring gastric emptying in this patient group and no further analyses were performed using this parameter(BT).

Strength of agreement between 13C-OBT and scintigraphic parameters

Based on linear regression, the strength of agreement between 13C-OBT and scintigraphy was modest. As demonstrated in Table 2, the strength of GEC in predicting gastric retention at all time points varied significantly depending on whether the prediction was for the group or an individual. At the mean GEC value, predictions of gastric emptying for the group ranged from within 6–9 % but the accuracy decreased to 34–50 % when estimating gastric emptying of individuals (Fig. 2). The best agreement between the methods was found when comparing the breath test GEC to scintigraphic retention at 120 min (r = −0.83; P < 0.0001).

Table 2 Details on the strength of agreement between breath test GEC and scintigraphic meal retention in predicting gastric emptying function for group and an individual
Full size table
Fig. 2
figure 2

Linear regression analyses, with 95 % confidence interval (95 % CI), to show the strength of agreement between breath test and scintigraphy for GEC in predicting gastric retention for an individual at a 60 min, b 120 min, c 180 min and d 240 min

Full size image

Cut-off GEC value for predicting delayed gastric emptying

On ROC analysis, the best cut-off GEC for predicting delayed gastric emptying (defined by scintigraphy as >13 % meal retention at 180 min [25]) was 3.25, with an AUC of 0.76 (95 % CI = 0.52–0.99; P = 0.054). When compared with scintigraphy, GEC <3.25 had a sensitivity of 89 %, a specificity of 67 %, a positive predictive value (PPV) of 92 % and a negative predictive value (NPV) of 57 % as a marker of delayed gastric emptying. The overall accuracy of breath test in assessing gastric emptying during critical illness was 86 %.

Relationship between GE and feed intolerance

Based on scintigraphy, delayed gastric emptying was found in all (100 %) patients with feed intolerance and 4/10 (40 %) patients with feed tolerance, giving 85 % sensitivity, 100 % specificity and 88 % accuracy of gastric scintigraphy for predicting feed intolerance. Conversely, using scintigraphy as the gold-standard, the cut-off GRV ≥250 ml to define feed intolerance was highly predictive (100 %) of the presence of delayed GE.

Based on 13C-OBT, delayed gastric emptying (GEC < 3.25) was present in 21/23 (91 %) patients with feed intolerance and 3/10 (30 %) patients with feed tolerance, giving a 88 % sensitivity, 78 % specificity and 85 % accuracy of 13C-OBT for predicting feed intolerance.

Discussion

This is the first study to formally compare gastric emptying measurements using the 13C-OBT and scintigraphy in critically ill patients. The results indicate that GEC correlates better with scintigraphy than t50(BT). The GEC also gives a reasonable estimate of gastric emptying, when used in a group setting, as may occur for research purposes. The 95 % confidence intervals, however, are relatively wide when the breath test is used to predict the gastric emptying of an individual. There is a close relationship between feed intolerance and delayed GE and all patients with feed intolerance had delayed GE as assessed by scintigraphy. Alternatively, the presence of delayed gastric emptying, assessed from either gastric scintigraphy or 13C-OBT, was highly predictive of feed intolerance (85–99 % accuracy).

Due to their non-invasiveness and ease of use, carbon-labelled breath tests are widely used to measure gastric emptying and have been validated in healthy, diabetic, dyspeptic, and gastroparetic subjects [9, 28, 29]. While the 14C-OBT has been shown to correlate well with gastric scintigraphy in critically ill patients [25], the test has been superseded by the more commonly used 13C-OBT, which has never been formally validated against scintigraphy in this group. 13CO2 excretion after labelled octanoate ingestion is dependent on gastric emptying, but it is unclear whether factors that are commonly present during critical illness, such as reduced intestinal absorption and altered hepatic function, can influence the absorption and/or metabolism of 13C octanoate and thereby the accuracy of the test.

As observed with 14C-OBT, both t50(BT) and GEC of 13C-OBT correlated significantly with gastric scintigraphic retention. However, the strength of correlation between the t50(BT) of the 13C-OBT (r = 0.55–0.66), was less than that of the 14C-OBT (r = 0.70–0.83). The reason for this discrepancy is unclear but may relate to the greater proportion of patients in our cohort with severe delayed gastric emptying, with extreme prolongation of t50(BT) (400 and 600 min) in five patients. It is noteworthy that the accuracy of the OBT has not been well studied at very slow rates of gastric emptying. In contrast, the performance of GEC is superior than that of t50(BT) with less variation, and this is observed in both 13C-based (r = −0.74 to −0.83) as in the current study and the 14C-based (r = −0.66 to −0.81) breath test as in the Chapman et al. study [25]. Although both variables were mathematically derived from the 13CO2 excretion curve, the GEC calculation accounts for both the appearance and disappearance of the label and was derived directly from a curve fitted to the raw data [9]. On the other hand, t50(BT) was calculated indirectly from the area under the observed and subsequent extrapolated curve [9, 27], which may explain its lesser accuracy and poorer correlation with scintigraphic meal retention. Our findings and those of others [1, 27, 30] suggest that GEC is the preferred breath test variable for assessment of gastric emptying in critically ill patients.

When measuring gastric emptying for clinical or research purposes, it is often helpful to be able to identify patients with slow gastric emptying. Data from a previous study where scintigraphic measurements of gastric emptying were performed in a group of healthy subjects suggested that gastric emptying was slow if there was >13 % retention at 180 min [25]. Using these criteria, our ROC analysis suggests a GEC of less than 3.25 may be the best cut-off value to indicate the presence of slow gastric emptying. Based on this definition, the 13C-OBT is moderately accurate (86 %) in identifying slow gastric emptying in critically ill patients with a 89 % sensitivity, 67 % specificity, 92 % PPV and 57 % NVP. These results are similar to those from 14C-OBT, which showed that a GEC of 3.03 gave the best cut off value for predicting delayed gastric emptying in critically ill patients, with a sensitivity of 75 % and specificity of 77 % [25]. We have also demonstrated good intra-subject repeatability with this technique [24]. Together, these data suggest that carbon labelled OBT is a relatively robust non-invasive and non-radioactive test in assessing gastric emptying in critically ill patients.

As illustrated in our study, the accuracy of the 13C-OBT in the assessment of gastric emptying in critically ill patients is lower than that in healthy subjects [9], in particular, with lower specificity. Whilst gastric scintigraphy solely measures the rate of gastric emptying, the results of the breath test technique are influenced by a number of factors other than gastric emptying, including intestinal absorption, liver metabolism and respiratory excretion of the labelled-carbon marker [9]. In healthy humans, gastric emptying is the rate-limiting step in the breath test technique as the other physiological parameters are normal. This may not be the case during critical illness as disturbances in small intestinal integrity and absorption [31, 32], liver function and lung function occur frequently [3]. These derangements may interfere with absorption and/or metabolism of 13C labelled octanoate and could be responsible for the lower accuracy of 13C-OBT in critically ill patients as compared to health. Furthermore, although pancreatic function has been reported to be reduced during critical illness [33], its impact on the absorption of octanoate is not known and warrants further evaluation. In addition, it is important to be aware that scintigraphy and the breath test are measuring two slightly different parameters with different units. Scintigraphy is presented in terms of percentage retention at certain times after instillation of the meal while the breath test is measured in terms of a half emptying time (minutes) or a calculated coefficient, the latter quantifying gastric emptying over the whole 4 h period. The two techniques cannot, therefore, be compared using the Altmann-Bland technique; and it is perhaps not surprising that the two techniques are not precisely comparable.

The borderline statistical significance on ROC analysis in the current study is most likely related to the relatively small sample size. Based on a power calculation (with 80 % power; α = 0.05; two tail), a sample size of at least 150 would be required to identify a normal range and allow categorical definition of delayed gastric emptying from the breath test. Even with such a large sample size, the estimated precision of the specificity for identifying delayed gastric emptying from breath testing remains only modest with a wide confidence interval (45–88 %). Thus, it is not a feasible exercise given the technical difficulties and tremendous efforts in conducting gastric scintigraphy in a large cohort of critically ill patients, for which the results may not provide a conclusive answer.

In addition to the relatively small sample size, the prevalence of delayed GE in our study appeared to be higher than would be expected from a random sample of critically ill patients [25, 27]. The higher prevalence of delayed GE in our study is most likely related to the increased severity of critical illness reflected by both the higher APACHE II score and the large number of “high risk admission diagnoses” such as head injury (43 %), multi-trauma (32 %) and sepsis (32 %). Available data indicate that disease severity is a major risk of delayed GE in critically ill patients. Another potential weakness of the current study is the definition of feed intolerance, which was assessed by GRV, and its reference as an indirect marker of delayed gastric emptying. Although our cut-off GRV >250 ml had been previously used by other studies that examined the issue of feed intolerance [5, 34], increasing data suggest that the relationship between GRV and gastric emptying is far from perfect and cautious interpretation must be adopted so that enteral feed will not be interrupted unless the GRV is greater than 500 ml [35].

In conclusion, the 13C-OBT is a useful technique for the measurement of gastric emptying in a group setting. The test therefore, remains an option for the measurement of gastric emptying in the critically ill for research purposes, for example to examine the effect of an intervention because it is convenient, non-invasive, has no radiation exposure and is technically simple. Using a GEC 3.25 as a cut-off, 13C-OBT has a reasonable accuracy in predicting the presence of delayed GE (86 %) but the modest accuracy in predicting gastric emptying in an individual coupled with relatively wide 95 % confidence intervals, make the test less useful in a clinical setting.