Abstract
Background/Aim: Cystoscopy, the standard diagnostic for bladder tumors, is uncomfortable, invasive, and expensive. The available urine-based marker systems all lack accuracy. Measuring volatile organic compounds (VOCs) from urine is a promising alternative. This pilot study evaluates the feasibility of discriminating bladder cancer patients' urine from healthy controls with an electronic nose. Materials and Methods: Headspace measurements of urine samples of 30 patients with confirmed transitional cell carcinoma (TCC) and 30 healthy controls were performed with Cyranose 320 calculating Mahalanobis distance and linear discriminant analysis. Histology reports following TUR-BT were correlated with urine findings. Results: After storage at −20°C, Cyranose correctly detected 28/30 already confirmed TCC samples and 26/30 healthy controls (p<0.01), sensitivity 93.3%, specificity 86.7%. Storage at −80°C led to similar results: 28/30 tumor samples and 28/30 control samples were correctly allocated; sensitivity and specificity both 93.3%. Conclusion: VOC detection is a promising tool to detect bladder tumors. Further research will test against possible confounders like bacteriuria.
About 20% of patients with visible haematuria and up to 5% of patients with microscopic haematuria have bladder tumors (1). In Germany, 28,500 patients were newly diagnosed with a bladder tumor in 2010 (2). Assuming a population of 80 million, this gives an incidence rate of 35.6/100,000 (3). 80% of them present with haematuria (1).
Most patients are diagnosed in an early non-muscle invasive stage (Ta or T1) and can undergo curative transurethral resection (TUR-BT), however they foster a significant risk of tumor recurrence and need careful follow-up.
The gold standard for the diagnosis of bladder tumors is cystoscopy, an uncomfortable, invasive, and expensive investigation (4). The fact that the workup of haematuria as well as the follow-up of non-muscle invasive bladder tumors require cystoscopy not only results in a great amount of patient discomfort, but also in significant financial impact on health systems. Over the past years, a multitude of urine-based biomarkers was introduced but none of them is a reliable substitute for cystoscopy and they are not used in routine daily practice (5).
The measurement of volatile organic compounds (VOC) is gaining importance in the field of tumor detection and is a promising approach towards the diagnostics of bladder tumors.
It has been established that many diseases leave a metabolic footprint: Altered enzymatic activity can be traced by the detection of the resulting chemical products in patients' breath, sweat, faeces, or urine. The idea of detecting the volatile components of these metabolic products and thus “smelling” diseases was first introduced by Nobel laureate Linus Pauling in 1971 (6). Different detection systems have since been evaluated: gas chromatography and mass spectrometry (GC-MS), selected ion flow tube and mass spectrometry (SIFT-MS), ion mobility spectrometry (IMS), and electronic noses which use a set of chemically differing sensors that detect changes of electrical resistance when a substrate from the analytical probe binds to them (7-9).
It could be shown that the detection of a range of pulmonary, neurological, and malignant diseases from breath is feasible by measuring VOCs (10-15). Furthermore, bowel diseases can be detected by headspace measurements of faeces (16-18).
Interestingly, the first studies on VOC-based bladder tumor detection in the headspace of urine investigated the use of trained sniffer dogs and could show a sensitivity of up to 73% (19).
The abovementioned technical detection systems could also successfully be employed to detect several tumor diseases from urine with variable accuracy. For some, the exact chemical alterations were tracked using solid phase microextraction and GC, and quantitative MS (20). Weber et al. used an electronic nose with 12 metal oxide sensors and 10 metal oxide semi-conductor field effect transistor sensors to detect bladder cancer and achieved 70% sensitivity and specificity (21). Khalid et al. established a different purpose-built detector system, based on GC-MS measurements with better accuracy and were also able, to some extent, diagnose prostate cancer (22, 23).
In this study, we successfully examined the feasibility of reliable bladder tumor detection with a commonly used electronic nose which has previously been established to detect pulmonary and neurological disease from breath (10, 11, 13).
Materials and Methods
Patients and sample storage. Fourty-two patients with cystoscopically confirmed bladder tumors referred for TUR-BT by their community Urologist were recruited from our outpatient department on the day of their pre-assessment. Patients with urinary tract infection on urine dipstick, bladder catheters or ureteric stents, and patients with known other malignancies were excluded. Twelve of those patients were excluded as no malignancy was found on histopathological workup. Thirty individuals with no known disease of the urinary tract were recruited as controls. Again, urinary tract infections were ruled out by dipstick.
All patients handed in a sample of first void morning urine for analysis which was split into aliquots of 2 ml for storage at −20°C and −80°C. Food and beverage consumption over the last twelve hours before specimen collection were recorded in detail using a standardised questionnaire.
The urine samples were obtained from the patients within three hours and immediately transferred to a fridge before aliquotation and freezing on the same day. Storage time until measurement was up to six months.
Ethics approval for this study was given by the local ethics committee (Az 131/14). All participants were informed about the study and handed an information sheet; written consent was obtained. The demographics of study participants are summarized in Table I.
Headspace measurements. Measurements were performed in small batches over a period of three months. Samples from both −20°C and −80°C stored in 2 ml vials were allowed to thaw at room temperature and then vortexed and heated to 37°C in a water bath for ten minutes before further analysis. Measurement of VOCs was performed in the open headspace of each sample using the Cyranose 320 electronic nose (Sensigent, Baldwin Park CA, USA). Every sample underwent two separate measurements.
The device contains 32 composite polymer sensors. An insulating polymer is combined with carbon nanospheres to create a semi-conductor. Upon exposure to the sample mixture, the volatile components are absorbed by the polymer resulting in swelling and consecutive reduction of conductive tracks which leads to an increase of resistance. Consequently, this generates a pattern according to the chemical composition of the VOC mixture.
Patient demographics.
The electronic nose training is done by measurement of a data set of well-defined samples in a first step. Based on the training set, the nose detects the unknown sample set as previously described (11). We have used the sensor systems for several years and could not detect sensor drift so far, regular tests are run by the distributing company without any hints for sensor drifting.
After warm-up of the device and calibration with an identification run, each measurement with the Cyranose 320 was performed at room temperature with the sample fresh from a 37°C water bath and consisted of three steps. 1. Baseline: Sensors were exposed to medicinal reference air (humidified with distilled water, flow at 1 l/min) for 60 sec. 2. Sampling: Sensors were exposed to sample air which was sucked in by a metallic syringe (“snout”) through a rubber plug neutral to the measurement for 60 sec. 3. Purging: Sensors were refreshed by exposition to medicinal reference air for 10 sec followed by ambient air for 50 sec.
Data analysis. The data were analysed as described before (10). In short, linear discriminant analysis (LDA) was used to distinguish between groups. As a variance-dependent distance measure for multidimensional data, Mahalanobis distance (MD) between groups was used.
A k-fold cross-validation was performed in each run, in order to calculate the cross-validation value (CVV), in which one data sample of each group was left out, and k= n1 · n2, with n1 and n2 being the sample sizes of groups 1 and 2. Every possible permutation, which holds one data sample of each group as the test set and the remaining data samples as the training set, was calculated. The classifications of the data samples of the test set were predicted via the corresponding training set, leading to the percentage of correct predictions, i.e. the CVV.
Results
On histopathological analysis after TUR-BT, transitional cell carcinoma was verified in 30/42 patients in the tumour group: 18 pTa, 4 pT1, 4 pT2 tumours, 2 Carcinoma in situ (CIS), 1 pT1 + CIS, and 1 pT2 + CIS (UICC 2010). Tumor differentiation showed high grade disease in 13/28 patients (excluding CIS) and low grade tumor in 15/28 (WHO 2016).
After storage at −20°C, linear discriminant analysis correctly allocated 28/30 patients to a “tumor” VOC pattern. Also, 26/30 healthy controls were correctly identified (p<0.001). Sensitivity to detect bladder cancer using the electronic nose was 93.3%, specificity reached 86.7%.
LDA for samples stored at −20°C.
The samples stored at −80°C did not harvest significantly different results: 28/30 tumor samples and 28/30 control samples were correctly allocated, resulting in a sensitivity and specificity of 93.3% each. Figure 1 shows the LDA data for −20°C, Figure 2 for −80°C.
A Mahalanobis distance of 1.58 between the two groups was calculated at −20°C (1.49 at −80°C) which displays a very good discrimination of both groups. CVV was 55% at −20°C and 53% at −80°C. No significant differences in measurement patterns could be identified between high grade and low grade tumors or for different T stages. Results for smokers and non-smokers did not differ.
Discussion
These first data are very promising and show that a commercially available electronic nose system can accurately distinguish between urine samples of bladder tumor patients and healthy individuals.
It has been shown that freezing and thawing of samples does not influence the VOC pattern (24). We could show that the storage temperature (−20°C vs. −80°C) did not significantly change results, either.
The ability of the electronic nose to detect bladder tumors was not affected by patients' smoking status, prior food consumption, or the presence of microscopic haematuria.
The Cyranose 320 electronic nose showed good diagnostic accuracy resulting in high sensitivity and specificity of 93.3% and 86.7% respectively after sample storage at −20°C. After storage at −80°C, both sensitivity and specificity reached 93.3%. A previous study with a different electronic nose showed 70% sensitivity and specificity (21). The device employed in this study has a larger array of sensors which may explain the higher diagnostic precision.
It seems that this setup may be equally efficient as a purpose-built detector system using gas chromatography for VOC separation previously described by Khalid et al. (22). Thus, chromatographic separation prior to VOC detection may represent an avoidable step for accurate measurements. Also, no pH adjustments were performed in this study.
LDA for samples stored at −80°C.
Our study is clearly limited by the small number of samples as well as the relatively strict inclusion and exclusion criteria. However, the preliminary results warrant further evaluation of the method in a larger cohort of patients presenting with visible haematuria for a full diagnostic workup.
In perspective, further research with larger numbers of participants is needed to establish whether the electronic nose can distinguish tumor grading and local tumor stage and whether it can be a means of non-invasive screening for bladder tumors in a population at risk (e.g. smokers) in the future. The system also needs to be validated against possible confounders such as urinary tract infection or visible haematuria. However, preliminary data from other centres with different VOC detection devices indicate that these factors are unlikely to influence the results (22).
It is yet unclear whether the urine composition of bladder tumor patients changes to a pattern similar to that of healthy controls after having their tumor resected.
Individual volatile markers can potentially be identified by gas chromatography and mass spectrometry in order to develop a further simplified testing device.
If the promising results of this pilot study are confirmed in a larger cohort, cystoscopy may potentially be avoided in the diagnostic workup for haematuria and in the follow-up for non-muscle invasive bladder cancer in the future.
In conclusion, the detection of volatile organic compounds with Cyranose 320, a well-established electronic nose system, is a promising approach for reliable non-invasive diagnosis and potentially for the follow-up of bladder tumors.
Acknowledgements
The Authors would like to acknowledge the support of Ursula Boas, Ecatarina Oplesch, and Helga Kirchner in the process of sample handling and measurements.
- Received November 23, 2017.
- Revision received December 12, 2017.
- Accepted December 13, 2017.
- Copyright© 2018, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.