Introduction

Airways surface liquid—the thin layer of fluid that overlies the airway epithelium—is critical for lung defence. The volume and composition of this liquid are tightly regulated and appear to determine its function. In animal models glucose is barely detectable in airways surface liquid [1]. The presence of sodium-glucose cotransporters in airways epithelial cells, capable of reabsorption of glucose, suggests that this low glucose concentration is actively maintained [2, 3]. Glucose in airways surface liquid could predispose to pulmonary infections by acting as a metabolic energy source for bacterial growth, by impairing host immunity through glycosylation of innate or acquired immune proteins or by reducing epithelial resistance to bacterial invasion through glycosylation of epithelial cells. Low glucose concentrations in airways surface liquid may therefore form an important part of lung defence against infection.

Few data exist on the glucose concentration of human airways secretions. Observations that nasal secretions did not contain glucose led to the use of glucose oxidase strip testing to distinguish between sugar-free nasal secretions and sugary cerebrospinal fluid (CSF) as a cause of rhinorrhoea in patients with suspected skull fractures. However, glucose was subsequently detected in nasal discharge in children without CSF leak [4]. Similar controversy has existed regarding glucose identified in endotracheal secretions of critically ill patients [5, 6]. The presence of glucose has been attributed to aspiration of enteral feed and used as a bedside test for aspiration by some investigators [5, 7], but others have disputed this use, suggesting the presence of glucose is independent of enteral nutrition [6].

Despite these studies, the normal concentration of glucose in secretions from healthy airways remains unclear and factors that could alter airways glucose have not been studied. Using the nose as an accessible area of upper airways epithelium, we aimed to establish whether glucose was present in airways secretions of people with healthy nasal epithelium. We then determined whether changes in the nasal epithelium (acute infective rhinitis) or in plasma glucose (diabetes) altered the levels of glucose present in nasal secretions. Finally, we determined the relationship between glucose concentrations in plasma and lower airways secretions in patients on an intensive care unit.

Materials and methods

Study groups

Subjects with healthy nasal mucosa or colds who were not diabetic were recruited from staff and students of St George's Hospital Medical School. Colds were defined as acute onset (in the last 10 days) of two or more of nasal discharge, nasal congestion, sneezing, cough, sore throat, fever or muscle aches. Subjects with a physician-made diagnosis of diabetes mellitus, irrespective of type of diabetes mellitus or treatment mode, were recruited from diabetic follow-up clinics at St George's Hospital. Subjects were not included if taking eye drops or nasal sprays. Patients admitted to a general adult intensive care unit and expected to require intubation for more than 24 h were recruited, irrespective of diagnosis.

All research was carried out in accordance with the Declaration of Helsinki (1989) of the World Medical Association, and was approved by the Local Research Ethics Committee. Informed consent was obtained from all the volunteers.

Clinical measurements

Nasal glucose study

Age and smoking history (non-smoker, ex-smoker or current smoker) of the subjects were recorded. Subjects underwent nasal glucose measurement at least 1 h after intake of food, drink or cigarettes. A glucose oxidase strip (Roche Diagnostics, UK) was placed in contact with the inferior surface of the inferior turbinate under direct vision. The strip was left in position for 30 s, then removed. Coating of the glucose oxidase area of the strip with nasal secretions was verified by inspection and, if inadequate, the strip was discarded and the sample repeated. The glucose concentration was determined from the colour change read against a visual colour scale. Nasal glucose was measured once in healthy and diabetic subjects. Subjects with colds who had glucose in their nasal secretions at the first measurement were recalled for a second measurement 2-weeks later when symptoms had resolved. In diabetics, blood glucose was measured by glucose oxidase strips read automatically (Accu-chek, Roche) and glycosylated haemoglobin (HbA1C) by laboratory analysis, as part of their routine clinical assessment.

Intensive care study

On admission to the intensive care unit, arterial blood, endotracheal secretions, and nasal secretions were sampled simultaneously for the measurement of glucose concentrations. Arterial blood glucose was measured using the glucose analyser on the blood gas machine (ABL2000, Radiometer, Copenhagen).

Endotracheal secretions sampled in a sputum trap were visually inspected and were rejected if obviously blood-stained. A glucose oxidase reagent strip was dipped into the samples and read against a visual colour scale to determine endotracheal glucose concentrations. Glucose concentrations were measured in nasal secretions by the method described for the nasal glucose study. The decision whether or not to feed subjects enterally using a nasogastric tube, was made by a clinician, independent of this study. The presence or absence or enteral feed (Ensure or Ensure Plus), as well as duration and success of feeding at the time of the study was noted. Enteral nutrition was continuous except for a 4-h rest in the night and was considered successful if <200 ml of feed (regardless of hourly rate) was retrieved through the nasogastric tube by aspiration attempted every 4 h. Patients fed within the previous 4 h or who had a gastric residual volume >200 ml within 12 h of enteral nutrition, were considered at risk of aspiration and analysed within the enterally fed group. The glucose concentrations of the Ensure and Ensure Plus, tested using glucose oxidase reagent strips and the glucose analyser, were both between 2 mmol l−1 and 4 mmol l−1.

Validation of methods

Glucose oxidase strips were used to measure glucose concentrations in nasal secretions because they enabled access to the low-volume secretions in patients and volunteers with normal nasal epithelium. These strips have been shown to reflect blood glucose concentrations accurately [8] and have been used successfully to measure glucose in endotracheal aspirates utilising a glucometer [5]. However, we found the cover of the strips to be too patchy for automated analysis and validated our method by comparing the glucose concentrations in sputum measured by glucose oxidase strips with concentrations measured by a glucose oxidase glucose analyser (GM9D, Analox Instruments, London, UK). To use the glucose analyser we required a non-particulate fluid and to obtain this, 1–2 ml sputum was forced through a 0.45 μm filter (Sartorius, Goettingen, Germany) yielding a non-particulate filtrate. The samples were obtained fresh from endotracheal suction and tested for glucose using the strips and processed immediately for the analyser. In addition, in order to ensure the method was tested over a wide range of glucose concentrations, five negative sputum samples were mixed with known concentrations of glucose solution (0–15 mmol l−1) and tested for glucose using the strips and the glucose analyser. A total of 50 samples were compared using this technique. Intensive care staff, unaware of the purpose of the study, read the oxidase strips to reduce observer bias.

Statistical analysis

Group values are given as mean±standard error of mean (SEM) for parametric data and median (interquartile range) for non-parametric data. Two-tailed P values of less than 0.05 were considered significant.

In the nasal glucose study, differences in frequency of detection of glucose in nasal secretions of healthy subjects and subjects with either diabetes mellitus or viral rhinitis were tested using χ2 analysis. Values for nasal glucose concentrations were compared between the groups using the Kruskal -Wallis test.

In the intensive care study, differences between the blood glucose concentrations and clinical characteristics (continuous variables) for patients with and without endotracheal glucose were tested using a t-test for independent samples. Chi-squared analysis was used to compare clinical characteristics (categorical variables) for patients with and without endotracheal glucose. Chi-squared analysis was also used to compare the presence or absence of glucose between endotracheal and nasal secretions and to determine the relationship between enteral feed and the presence or absence of endotracheal glucose.

For the validation of methods, a Bland and Altman plot was constructed. The mean and standard deviation of the differences between methods was calculated and tested for significance using a one-sample t-test of the differences against zero.

Results

Nasal glucose study

Normal volunteers

Nineteen volunteers (nine male, ten female, age 23.3±1.2 years), with normal nasal mucosa and normal plasma glucose, underwent nasal glucose measurement. Sixteen were non-smokers and four were ex-smokers. Glucose was not detected in nasal secretions from any of the normal volunteers (Fig. 1).

Fig. 1.
figure 1

Glucose concentrations in nasal secretions of healthy volunteers, volunteers with common colds, and patients with diabetes mellitus

Full size image

Rhinitis

Twenty-four individuals (eight male, 16 female, age 22.3±0.6 years), with nasal epithelial inflammation due to acute viral rhinitis, were studied. Nineteen were non-smokers, one was an ex-smoker, and three were current smokers. Glucose [1 (IQR, 1–2) mmol l−1] was detected in the nasal secretions of 12 (50%) subjects (χ2=13.2, 1 df, P<0.0001 vs healthy controls, Fig. 1), but was not related to smoking habit. Ten of the people with detectable nasal glucose were re-tested 2-weeks later, nine had no nasal glucose detected at the second measurement. The one person who still had 1 mmol l−1 glucose in nasal secretions at repeat measurement had persistent coryzal symptoms.

Diabetes mellitus

Twenty subjects (nine male, 11 female, age 58.4±3.4 years) with diabetes mellitus, but normal nasal mucosa, were studied. Fourteen were non-smokers, four were ex-smokers and two were current smokers. Glucose [4 (IQR, 2–7) mmol l−1] (Fig. 1) was detected in the nasal secretions of 18 patients with diabetes mellitus (χ2=31.8, 1 df, P<0.0001 vs healthy controls). Only two people with diabetes did not have glucose in their nasal secretions. These individuals had good glycaemic control as measured by HBA1C and both had plasma glucose <8 mmol l−1 at the time of measurement.

Intensive care study

Sixty patients were studied (25 female, 35 male, mean age 59±2.2 years). Forty-four patients had random blood glucose greater than 7 mmol l−1 during their ICU admission.

Relationship between glucose concentrations in blood and airways secretions

Figure 2 illustrates the relationship between the highest blood glucose obtained for each patient and the glucose concentration measured simultaneously in endotracheal secretions. Thirty-one patients had glucose detected in endotracheal secretions (blood glucose 9.8±0.4 mmol l−1), including all patients with a blood glucose concentration greater than 10.1 mmol l−1. Twenty-nine patients had no glucose in their endotracheal secretions (blood glucose 7.2±0.3 mmol l−1) and of the 16 patients with a blood glucose less than 7.0 mmol l−1, only four had endotracheal glucose. Table 1 describes the demographics of patient with and without, endotracheal glucose. There were no significant differences in underlying diagnosis or illness severity between the two groups.

Fig. 2.
figure 2

Comparison of simultaneous measurements of glucose concentrations in arterial blood and endotracheal secretions from patients on a general intensive care unit

Full size image
Table 1. Characteristics of patients with and without glucose in endotracheal secretions
Full size table

In 42 patients both endotracheal and nasal glucose were measured. There was strong concordance in the presence or absence of glucose between nasal and endotracheal secretions (Table 2, P<0.001).

Table 2. Relationship between the presence and absence of glucose in endotracheal and nasal secretions
Full size table

Relationship between enteral feeding and presence or absence of endotracheal glucose

Endotracheal glucose was detected in 18 out of 32 patients receiving enteral feed at the time of investigation and 13 out of 28 patients who were not receiving enteral feed (P = 0.309, χ2). There was no difference in the concentration of glucose measured in endotracheal secretions of patients receiving enteral feed (concentration 3.7±0.8 mmol l−1) and those not receiving feed (concentration 4.0±0.9 mmol l−1). Furthermore, there was no difference in the administration of glucose-containing oral medications between the patients with and without endotracheal glucose.

Validation of method

Figure 3 is a Bland and Altman plot comparing the use of a glucose oxidase strip and a glucose oxidase analyser for the measurement of glucose in sputum. The mean of the differences between the methods was 0.02±1.92 (not significant). For the glucose oxidase strips there were no false positive readings and five false negative readings.

Fig. 3.
figure 3

Bland-Altman plot comparing the use of glucose oxidase strips and a glucose oxidase analyser to measure glucose concentrations in sputum

Full size image

Discussion

Our study shows that glucose is not detectable in normal nasal secretions from people with healthy nasal epithelium and without diabetes mellitus using glucose oxidase reagent strips. These results are consistent with animal studies that have found airways surface liquid glucose concentrations to be 20-fold lower than those of plasma [1]. The mechanisms that maintain low glucose concentrations in airways secretions are not clear, but may include active glucose absorption from airways surface liquid by glucose transport proteins. In support of this, active glucose absorption from lumen to interstitium has been demonstrated in sheep whole lung [1]. Expression of sodium-glucose co-transporters, capable of sodium-driven glucose absorption against a glucose concentration gradient, has been demonstrated in human lung tissue [9].

We did detect glucose in nasal secretions from 90% of subjects with diabetes mellitus and 50% of those with nasal epithelial inflammation due to acute rhinitis. In both conditions disturbance of normal glucose absorption across airways epithelium could account for these findings. Hyperglycaemia could increase diffusion of glucose into airways secretions, exceeding the capacity for glucose reabsorption and rhinitis could reduce glucose reabsorption by disruption of nasal epithelial integrity. Alternatively, in rhinitis, the presence of glucose might be explained by transepithelial exudation of plasma secondary to inflammation.

Approximately half of the patients ventilated on ICU did not have glucose detected in their endotracheal secretions. The mean blood glucose concentration for these patients was lower than for patients with endotracheal glucose, and we hypothesise that the mechanisms, which exist to maintain the airways surface liquid devoid of glucose, were intact. However, glucose was present in the endotracheal secretions of 31 out of 60 patients studied on the intensive care unit. Observations in our non-ventilated groups suggest that the appearance of glucose in airways secretions could be secondary to either inflammation or hyperglycaemia. Patients with glucose in endotracheal secretions had higher blood glucose concentrations (9.8±0.4 mmol l−1) than those without endotracheal glucose (7.2±0.3 mmol l−1), and all patients with blood glucose >10.1 mmol l−1 had glucose in their endotracheal secretions. Hyperglycaemia is associated with the appearance of glucose in other bodily fluids. Glucose appears in the urine of type 1 and type 2 diabetics at blood glucose concentrations ranging from 6.0 mmol l−1 to 14.3 mmol l−1 and 6.2 mmol l−1 to 12.3 mmol l−1, respectively [10, 11]. In stimulated salivary secretions glucose appears at blood glucose concentrations between 10–15 mmol l−1[12]. Our observations in ICU patients suggest a similar relationship between blood glucose concentrations and the appearance of glucose in lower airways secretions.

Hyperglycaemia cannot account for the presence of glucose in the endotracheal secretions of patients with normal blood glucose concentrations (Fig. 2). However, our observations from the volunteers with acute rhinitis, suggest that the presence of glucose in the airways secretions of these patients could be attributable to airways inflammation. In this small observational study we have not been able to study the relationship between respiratory pathology and the presence or absence of glucose, although there were no differences in the severity of illness score between patients with, and without, endotracheal glucose (Table 1).

Detection of glucose in endotracheal secretions has previously been interpreted as a sign of aspiration of enteral feed. Winterbauer and colleagues [7] made repeat measurements of glucose in respiratory secretions from 20 critically ill patients receiving tube feedings, and related these to clinical signs of aspiration. These included tachypnoea, tachycardia, fever, hypoxaemia, and new infiltrate on chest radiograph, no other explanation for these findings, and resolution of syndrome coincident with discontinuation of feeding. They found that six patients with "clinically-significant aspiration" had glucose consistently detectable in respiratory secretions on repeated testing. Although they interpreted their results as indicative of the presence of enteral feed in the airways, it is possible that airways inflammation or changes in blood glucose could also have explained their findings. In our study we found glucose in the endotracheal secretions of a similar proportion of patients who were or were not receiving enteral feed. Furthermore there was no difference in the glucose concentration of endotracheal secretions in those who were receiving enteral feed (3.7±0.8 mmol l−1) and those not receiving feed (concentration 4.0±0.9 mmol l−1). These observations are similar to the findings of Kinsey and colleagues [6] who also found that endotracheal glucose concentrations were not different in enterally fed patients (3.7±3.0 mmol l–1) and non-fed patients (5.8±3.9 mmol l−1). In our patients who did receive enteral feed, the glucose contents of the enteral feeds (2–4mmol l−1) measured with glucose oxidase sticks were not sufficient to account for higher levels of endotracheal glucose (up to 11 mmol l−1, Fig. 2). Patients in this study were fed using either Ensure or Ensure plus, commercial enteral feeds, which in their UK formulation, contain the carbohydrate sources sucrose and maltodextrin. Neither of these carbohydrates reacts with glucose oxidase and there is little or no amylolysis of maltodextrins by salivary amylase to reactive dextrose [13].

It is important to consider the limitations of methods used in this study when interpreting our results. We used glucose oxidase strips to measure the glucose concentrations in airways secretions. We made nasal glucose measurements by direct application of a glucose oxidase strip to the nasal mucosa. To optimise the accuracy of this technique, coating of the glucose strip with nasal secretions was checked visually after removal of the strip from the nose. This safeguard ensured that where glucose was undetected, this was due to absence of glucose from the secretions rather than inadequate sampling. The sampling procedure was repeated if the strip remained dry. We chose to measure glucose in nasal secretions directly by this method rather than by stimulating and collecting secretions for laboratory analysis as all measures used to stimulate secretion change the nature of the secretions [14]. Glucose oxidase sticks have previously been evaluated against laboratory analysis for measuring glucose in endotracheal secretions [5] but are not designed this purpose. We therefore compared the endotracheal glucose measurements by our glucose oxidase strips with an automated glucose oxidase analyser. We found the strips reliably indicated the presence or absence of glucose and accurately measured low concentrations of glucose (<3 mmol l−1) in airways secretions. However, as the glucose concentration in airways secretions increased, the ability of the stick method to discern colour changes accurately became less reliable (Fig. 3). In the ICU study a number of different observers read the glucose oxidase sticks to reduce possible bias. It is possible that the ability to discern colour change might be improved by utilising one practised individual for the interpretation.

One potential criticism of our study is that we have extrapolated results of glucose measurements in the nasal secretions of volunteers to explain the presence of glucose in lower airways secretions of ventilated ICU patients. In the non-ventilated patients, we measured nasal glucose because nasal secretions were easily accessible and this avoided the need for invasive procedures such as bronchoscopy in volunteers. Although nasal and large airways mucosae and secretions are not identical, they do share many features including ciliated epithelial cells and submucosal glands, and have similar processes controlling the volume and composition of airways surface liquid, including ion transport proteins such as the epithelial sodium channel and cystic fibrosis transmembrane regulator. We found tight agreement in the presence or absence of glucose between nasal and lower airways secretions in ICU patients, suggesting that factors that alter nasal glucose concentrations may also determine lower airways glucose concentrations.

We have shown that glucose is not normally present in the nasal secretions of healthy volunteers, but that it becomes detectable where hyperglycaemia or epithelial inflammation occur. The detection of glucose in endotracheal secretions from patients intubated on intensive care is related to, but not completely accounted for by hyperglycaemia. Glucose can be detected in airways secretions from patients with and without enteral feeding and therefore cannot reliably be used to detect aspiration of enteral feed.