Prostate Cancer Diagnosis Through Electronic Nose in the Urine Headspace Setting: a Pilot Study

ORIGINAL ARTICLE Prostate cancer diagnosis through electronic nose in the urine headspace setting: a pilot study

AD Asimakopoulos1, D Del Fabbro2, R Miano2, M Santonico3, R Capuano4, G Pennazza3, A D’Amico4 and E Finazzi-Agro` 2

BACKGROUND: To evaluate the efficacy of prostate cancer (PCa) detection by the electronic nose (EN) on human urine samples. METHODS: Urine samples were obtained from candidates of prostate biopsy (PB). Exclusion criteria were a history of urothelial carcinoma or other malignant disease, urine infection, fasting for o12 h before PB or ingestion of alcohol or foods that might alter the urine smell in the last 24 h. The initial part of the voided urine and the midstream were collected separately in two sterile containers. Both samples were analyzed by the EN immediately after the collection. All patients underwent a standard transperineal, transrectal-ultrasound-guided PB. The pathological results were compared with the outcomes of the EN. Sensitivity and specificity of EN were assessed. RESULTS: Forty-one men were included in the study. Fourteen out of the 41 patients were positive for PCa. Midstream urine did not correlate significantly neither with a positive nor with a negative PB. Instead, significantly different results on the initial part of the urine stream between positive and negative PBs were obtained. The EN correctly recognized 10 out of the 14 cases (that is, sensitivity 71.4% (confidence interval (CI) 42–92%)) of PCa while four were false negatives. Moreover, the device recognized as negative 25 out of the 27 (that is, specificity 92.6% (CI 76–99%)) samples of negative PBs, with only two false positives. CONCLUSIONS: We believe this is the first demonstration of an olfactory imprinting of the initial part of the urine stream in patients with PCa that was revealed by an EN, with high specificity.

Prostate Cancer and Prostatic Disease (2014) 17, 206–211; doi:10.1038/pcan.2014.11; published online 1 April 2014

INTRODUCTION MATERIALS AND METHODS Prostate cancer (PCa) is the most frequent noncutaneous Urine samples were obtained from Caucasian patients recruited in our malignancy in men.1 Although PSA blood testing remains the center following written consent for analysis of their urine for research most widely used tool for PCa detection,2 it has been historically programs. All men included were consecutive candidates of PB. Our local characterized by lack of specificity3 while none of the novel urine ethics committee was informed of the study; a formal approval was considered unnecessary, because our observational preliminary study did or blood biomarkers—proposed in the past decade—is currently not change the normal clinical practice: the urine samples used for the EN widely used.4 5 evaluation were immediately dismissed and they were not used for any Recently, Matsumura et al., for the first time, established the other clinical or scientific purpose, while EN findings did not change the feasibility of using urinary volatiles to detect lung cancer. In the clinical behavior of the involved medical staff. field of PCa, Cornu et al.6 performed a double-blind study using A complete blood count, a basic metabolic panel (blood urea nitrogen, an adequately trained dog in order to check its ability to detect creatinine, glucose, electrolytes) and a urinanalysis, were required to all PCa by sniffing urine. The results were surprising, with the dog candidates to PB as standard clinical practice. correctly designating the cancer samples in 30 of the 33 cases, Exclusion criteria for our study were a history of urothelial carcinoma or conferring a 91% of both specificity and sensitivity to the test. other malignant disease, any urine infection, fasting for o12 h before PB or ingestion of foods that might alter the urine smell (that is, asparagus, Although issues about the reproducibility (significantly affected by cauliflower, garlic, meat) in the past 24 h. the canine olfactory receptor polymorphisms that influence odor 7 Each patient was given two sterile containers for urine collection, one for detection performance of sniffer dogs ), costs and duration of the the initial part of the voided urine and the other for the midstream. The method were raised, we believe this study provided the first urine sample collection was performed just before the PB, independently demonstration that PCa gives an odor signature to urine. of the fact that it was or not the first voided urine of that morning. Volatile Using the electronic nose (EN), Roine et al.8 recently proved that organic compound (VOC) analysis was performed before the PB and both malignant and nonmalignant prostate cell lines have distinct without knowing the PB results by medical and engineering staff not smell prints. involved in the PB procedure. Furthermore, the staff performing the PB Based on these findings, we herein present, to the best of our was not aware of the VOC analysis results. An adequately designed obturator was created in order to permit the knowledge, the first study on the outcomes of the EN in the extraction of the urine headspace necessary for the EN analysis (Figure 1). identification of PCa in urine samples of prostate biopsy (PB) Both samples were then analyzed by last version of the EN as developed at candidates. Our aim was to evaluate the performance of this the faculty of Engineering of the University of Rome ‘Tor Vergata’ device as a conventional, non-invasive, quick and easily available (Figure 2).9 All measurements were performed within two hours from the instrument that, added to PSA, may improve PCa diagnosis. collection of the urine samples, in order to maintain unaffected the

1UOC of Urology, Department of Surgery, University of Rome Tor Vergata–Policlinico Casilino, Rome, Italy; 2UOC of Urology, Department of Surgery, University of Tor Vergata, Policlinico Tor Vergata, Rome, Italy; 3Center for Integrated Research—CIR, Unit of Electronics for Sensor Systems, Universita` Campus Bio-Medico, Rome, Italy and 4Department of Electronic Engineering, University of Tor Vergata, Rome, Italy. Correspondence: Dr AD Asimakopoulos, UOC of Urology, Department of Surgery, University of Rome Tor Vergata– Policlinico Casilino, Viale Oxford 81, Rome, 00133, Italy. E-mail: [email protected] Received 7 November 2013; revised 15 January 2014; accepted 9 February 2014; published online 1 April 2014 Prostate cancer and electronic nose AD Asimakopoulos et al 207 cases or controls after pathological examination of the specimens. The peripheral zone of both lobes from the lateral to the paramedian area and from the base to the apex was sampled. For prostate volumes450 ml and on the basis of the clinical/ultrasound characteristics, additional cores were eventually obtained. A dedicated uropathologist examined the slides. Finally, the pathological results were compared with the outcomes of the EN. The overall graphic representation of the measurements obtained from the eight sensors requires an eight-dimensional space. A multivariate analysis obtained with supervised PLS-DA (Partial Least Square– Discriminant Analysis) permitted the graphic representation of the sensors’ measurements on a bi- or three-dimensional model. The PLS-DA model was built by using the leave-one-out as cross- validation criterion. All the statistical data relative to the model are reported in Table 1. All the statistical parameters of the sensor array data Figure 1. The adequately designed obturator created in order to set are reported in Figure 3. permit the extraction of the urine headspace, necessary for the The classiﬁcation model here presented was built using the PB electronic nose analysis. outcomes as reference diagnosis.

Statistical methods The kappa statistic (k) was used as a measure of chance-corrected agreement between PB and EN. k-values ofo0.20, 0.21–0.40, 0.41–0.60 E.N. and40.61 represent poor, fair, moderate and good agreement, respectively. Sensitivity, speciﬁcity, positive predictive value—negative predictive FILTERS value and their conﬁdence interval (CI) were calculated.

SOFTWARE RESULTS Fifty consecutive patients referring to our center for PB were evaluated for eventual analysis of their urine headspace through EN. By applying the exclusion criteria, nine candidates to PB were excluded from the evaluation with EN. Consequently, 41 men URINE aged 54–77 years were included in the study. Nine of them were SAMPLE smokers, seven were diabetics and 34 assumed oral medications for various pathologies. One patient was liver-transplanted. There was no statistically significant difference in the distribu- Figure 2. Electronic nose (E.N.). tion of smokers vs non-smokers, alcohol consumers vs non- alcohol consumers and between patients under oral medications or not between the control and cancer group (Table 2). PSA ranged between 1.95 and 14.02 ng ml À 1. A mean of 12.84 cores was obtained from the PBs (range 12–17). At the final histology, 14 olfactory characteristics of these samples. All the principles of the reference patients (34%) were positive for PCa; 12 of them had a Gleason standard (European Standard, EN 13725, Air quality–Determination of score of 6 (3 þ 3) and two a score of 7 (3 þ 4). odour concentration by dynamic olfactometry) that is used to provide a Midstream urine did not correlate significantly neither with a scientifically objective method for assessing odours were respected. The device used is equipped by eight non-selective gas sensors. Each positive nor with a negative PB; the outcomes on these samples sensor is coated with different metallo-porphyrins. Cu-Tetra Phenyl were consequently excluded from the final analysis. On the Porphyrin (TPP), Co-TPP, Zn-TPP, Mn-TPP, Fe-TPP, Sn-TPP, Ru-TPP and contrary, we obtained significantly different results on the initial Cr-TPP are the eight metals selected, showing a large affinity toward part of the urine stream between positive and negative PBs. multiple different VOCs. Each sensor (transducer þ metallo-porphyrin) The agreement between PB and EN was 85%, that is, adsorbs on its surface the VOCs and changes its frequency of resonance significantly higher than the expected by chance (Po0.001) with because of its mass variation as regulated by the proportional law of k 10 a value of 0.66. Sauerbrey. The interaction between the VOCs and the sensors is In particular, the EN correctly recognized 10 out of the 14 cases regulated by weak bonds, such as Van der Waals, dipole–dipole and (that is, sensitivity 71.4% (CI 42–92%)) of PCa while four were false hydrogen. The change on the frequency of each sensor with respect to the baseline constitutes the response of that sensor. The final output consists negatives. Both patients with Gleason 7 disease were correctly of a fingerprint of the eight sensor responses registered for each sample. recognized by the EN. More importantly, the device recognized as A matrix containing all the measurements performed is then extracted by negative 25 out of the 27 (that is, specificity 92.6% (CI 76–99%)) means of a dedicated software. samples of negative PBs, with only two false positives (Table 3). Two hundred seconds were required for the analysis of each sample The two false-positive patients were followed-up with digital (with a flow rate of 50 Standard cc3 per min) while 600 s of flow with rectum exam and PSA evaluation at 6 months after the initial dehumidified air were required between two consecutive samples in order biopsy. A rebiopsy was required to one of them for rising PSA, to clean the sensors. resulting again negative for PCa. All the components with which the urine samples and the urine Globally, the positive predictive value of the device is 83.3% (CI headspace get in contact during the analysis (teflon for the obturator and tygon for the tubes bringing the headspace to the sensors) are chemically 52–98%), while the negative predictive value is 86.2% (CI 68–96%). inert, odorless and they present a low potential of absorption towards the Our results are depicted in Figure 4. VOCs that remain consequently not affected. The mean number of positive bioptic cores among the correctly Subsequently, all patients underwent a transperineal, transrectal- identified PCa patients was 3.2 (range: 1 positive core with 10% ultrasound-guided PB according to a standard procedure (12 cores) under involvement of the core–7 positive cores with 10–80% involve- local anaesthesia with an 18-G biopsy needle. They were classified as ment of the cores).

& 2014 Macmillan Publishers Limited Prostate Cancer and Prostatic Disease (2014), 206 – 211 Prostate cancer and electronic nose AD Asimakopoulos et al 208 300 Table 1. Statistical parameters of the PLS-DA model calculated on the EN data, including the PRESS (cumulative prediction error sum of squares) 250

200

150 frequency shift [Hz] 100

0 1 2 3 4 5 6 7 8 sensors Figure 3. Boxplot resuming all the statistical parameters of the sensor array data set. On each box, the central mark is the median, the edges of the box are the 25th and 75th percentiles, the whiskers extend to the most extreme data points not considered outliers, and outliers are plotted individually.

DISCUSSION The early diagnosis of cancer constitutes a basic and diachronic objective of oncology. Animal models are increasingly used in this context. Canines are mainly adopted due to their superior olfactive apparatus, characterized by a detection threshold of parts per trillion. The scientific basis of this ability of dogs to detect the odor signature of cancer is linked to the VOCs produced by malignant cells.11 In fact, during tumor growth, protein changes in malignant cells lead to peroxidation of the cell membrane components and produce VOCs that can be detected in the headspace of the cells.12,13 The utilization of dogs for cancer detection merged after the first case report in 1989 about a melanoma detected by a dog on his owner’s leg.14 There followed another study on melanoma detection by dogs directly on tissue that was correct in 75–86% of the cases.15 Other studies documented an impressive performance of the canine olfatus in detecting lung cancer by sniffing the exhaled breath (sensitivity 71%, specificity 93%, independent of the presence of chronic obstructive pulmonary disorder or tobacco smoke),16 ovarian cancer in both blood and tissue samples (reaching a sensitivity of 100% and specificity of 98%),17 breast cancer in breath samples (sensitivity and specificity of dog detection of 88% and 98%, respectively),18 colorectal cancer (sensitivity of 91% and 97% in breath and fecal samples, respectively, and a specificity of 99% for both sample types)19 and bladder cancer by sniffing urine samples, with a diagnostic success rate of 41%.20 By sniffing urine, dogs can accurately detect lung cancer5 as On this basis, it was possible to calculate the correct number of latent well as bladder or breast cancer with better than chance variables (LVs) to be selected in order to avoid overfitting (least RMSECV: probability,21 although no positive result was obtained for PCa.22 root mean square error in cross validation). EN, electronic nose; PLS-DA, 6 partial least square-discriminant analysis. In the specific field of PCa, Cornu et al. showed for the first time that a trained dog can distinguish a PCa urine sample among controls with powerful results: the dog correctly designating the The mean number of positive bioptic cores among the four cancer samples in 30 of the 33 cases, conferring a 91% of both misclassified patients was 1.75 (range 1 positive core with 5% specificity and sensitivity to the test. The main success of this involvement–3 positive cores with 10–40% involved). No statisti- study was the demonstration that PCa gives an odor signature to cally significant difference between the mean number of positive the urine. cores between correctly identified PCa and misclassified patients However, these results were obtained with only one dog, was observed. questioning the reproducibility of this method. Furthermore, the Figure 5 represents the receiver operating characteristic curve type of dog used in the study could have influenced the results as for the discrimination between positive and negative subjects. canine olfactory receptor polymorphisms have been shown to

Prostate Cancer and Prostatic Disease (2014), 206 – 211 & 2014 Macmillan Publishers Limited Prostate cancer and electronic nose AD Asimakopoulos et al 209 Table 2. Effect of the confounding factors on the outcomes of the EN

Negative Positive Marginal row totals

Smoking 6 (5.93) [0] 3 (3.07) [0] 9 Non-smoking 21 (21.07) [0] 11(10.93) [0] 32 Marginal column totals 27 14 41 (grand total)a Alcohol 6 (7.24) [0.21] 5 (3.76) [0.41] 11 Non-alcohol 21 (19.76) [0.08] 9 (10.24) [0.15] 30 Marginal column totals 27 14 41 (grand total)b Pharma 22 (22.39) [0.01] 12 (11.61) [0.01] 34 Non-pharma 5 (4.61) [0.03] 2 (10.24) [0.06] 7 Marginal column totals 27 14 41 (grand total)c Abbreviation: EN, electronic nose. aThe w2 statistic is 0.003. The P-value is 0.953574. This result is not significant at Po0.05. bThe w2 statistic is 0.8549. The P-value is 0.355168. This result is not significant at Po0.05. cThe w2 statistic is 0.1167. The P-value is 0.732676. This result is not significant at Po0.05.

Table 3. Suggested areas to focus the future research on the EN

Abbreviations: EN, electronic nose; PCa, prostate cancer; VOC, volatile organic compound.

influence odor detection performance by sniffer dogs.7 The study gas concentration in real time. The operating principle of the was costly and long in duration, making it difficult to conceive of piezoelectric chemical sensors (also defined mass sensors) located an extended use for this test in clinical practice. Finally, concerns in the EN narici, is based on the variation of frequency of regarding the lack of communication with the dogs, the difficulties resonating quartz crystals (Thickness Shear Mode Resonators) of an adequate training, several hygienic and even ethical issues caused by their mass variation, that its turn is determined by the regarding dogs subjected to forced inhalation should be further adsorption or desorption processes of gas molecules on the considered. chemical interactive materials (metal-tetrafenilporfirine) of the Although canine cancer detection is not feasible for wide- sensors. The odour classification is obtained by a kind of spread clinical use, these studies have inspired a race between multivariate statistical analysis. scientists to develop suitably sensitive analytical instrumentation The interaction between VOC with an array of sensors generates for quick and easy clinical use that show similar accuracy to the a characteristic fingerprint, which can be recognized by compar- mammalian sense of smell.23 By the term EN it is intended an ing it with previously recorded patterns in the recognition system. instrument able to generate digital maps of complex odours or The common features of all ENs are the relatively fast response chemical images. The fundamental elements of the EN that we and the satisfactory stability. These devices are used in the food used are the piezoelectric chemical sensors which operating as and agricultural sectors, in the ambient monitoring, in space miniaturized transducers do respond reversely to the chemical applications and for security issues. In the medical field, they are volatile compounds generating electrical signals in function of the used as diagnostic instruments for skin disease and internal

& 2014 Macmillan Publishers Limited Prostate Cancer and Prostatic Disease (2014), 206 – 211 Prostate cancer and electronic nose AD Asimakopoulos et al 210 Scores Plot Loads Plot 4 0.6

3 0.5

2 0.4 1 Controls 0.3 0

0.2 −1 LV 1 (38.55%) LV 2 (25.01%)

−2 0.1 Prostate cancer −3 0 −4 −0.1 −5 1 2 3 4 5 6 7 8 −6 −4 −2 0 2 4 6 Variable Number LV 1 (38.55%) Loads Plot Figure 4. Scores plot of the latent variables (LV) 1 and 2 obtained by 0.3 the PLS-DA (Partial Least Square–Discriminant Analysis) model. It is evident the discrimination between control subjects (label 0) and 0.2 tumor cases (label 1). 0.1

−0.1 1.00 −0.2 LV 2 (25.01%)

0.75 −0.3

−0.4

0.50 −0.5 Sensitivity −0.6 1 2 3 4 5 6 7 8 0.25 Variable Number Figure 6. Loading plots for the latent variables (LV)1 of the PLS-DA (Partial Least Square–Discriminant Analysis) model calculated on the 0.00 electronic nose data. Non-negligible load values can be observed, 0.00 0.25 0.50 0.75 1.00 meaning that all the sensors give an important contribution to the 1 - Specificity patient’s discrimination. Area under ROC curve = 0.8915 Figure 5. Receiver operating characteristic (ROC) curve with line of identity of breathprint latent variable 1 predictive for prostate prints of collected prostatic nonmalignant (EP-156 T and controls) cancer diagnosis. (area under the ROC curve 0.8915). and malignant (LNCaP) cell lines. The device differentiated the nonmalignant and malignant cell lines from each other with very low misclassification rates (2.9–3.6%), suggesting that malignant and nonmalignant cell lines have distinct smell prints.8 pathology detection, such as diabetes, cancers, urinary infections, With our study, we demonstrated, we believe for the first time, tuberculosis, Helicobacter pylori and so on.21,24 in the clinical setting (and not in cell lines) a potential role of the Focusing on the field of oncology, dynamic olfactometry, EN in identifying, by smelling urine samples, healthy patients obtained with EN, is currently studied as an alternative to the (with negative PB) while maintaining an acceptable sensitivity. The animal model for the early detection of several human tumors. overall agreement between biopsy and EN was 85% while Successful detection of lung,25–30 gastric,31 ovarian,32 colorectal,33 the k value (0.66) indicates a good agreement between the breast18,34 and skin cancer35 by EN are well described in the two methods. published literature, with results similar, even if somewhat lower, The first part of the urine stream (and not the midstream) to those of the dogs. allowed to distinguish between positive and negative PBs, In the field of urological oncology, the EN has already shown probably because it contains elements of prostatic secretion. promising results concerning the detection of urinary tract tumors The influence of eventual confounding factors that might alter in urine samples. EN was able to discriminate urine samples the urine smell influencing the EN accuracy has been also of healthy patients from those of patients with bladder cancer, evaluated. Alcohol, oral medications and smoke do not seem to with a diagnostic accuracy of 100%,36 overcoming the one influence the accuracy of the device. described for dogs.20 Overall, the results obtainable with the EN are chemical images The diagnostic accuracy of EN in PCa was not verified with a rather high probability of cancer recognition. The high till recently, when Roine et al.8 analyzed through EN the smell specificity of the device, if confirmed, could allow to reduce the

Prostate Cancer and Prostatic Disease (2014), 206 – 211 & 2014 Macmillan Publishers Limited Prostate cancer and electronic nose AD Asimakopoulos et al 211 number of first PBs or rebiopsies. It would appear that if the test 7 Lesniak A, Walczak M, Jezierski T, Sacharczuk M, Gawkowski M, Jaszczak K. results negative, the PB will be almost always (93%) negative; thus, Canine olfactory receptor gene polymorphism and its relation to odor detection both extra costs to the health-care system and potential performance by sniffer dogs. J Hered 2008; 99: 518–527. complications to the patients of further diagnostic procedures 8 Roine A, Tolvanen M, Sipilaïnen M, Kumpulainen P, Helenius MA, Lehtima¨ki T et al. may be spared. Detection of smell print differences between nonmalignant and malignant Two issues regarding the sensors should be further commented: prostate cells with an electronic nose. Future Oncol 2012; 8: 1157–1165. 9 Santonico M, Pennazza G, Capuano R, Falconi C, Vink TJ, Knobel HH et al. their statistical contribution and their consistency with respect Electronic noses calibration procedure in the context of a multicentre medical to prostate disease-alleged biomarkers. It is worth mentioning study. Sens Actuators B Cheml 2012; 173: 555–561. that the used EN consists of an array of non-selective gas sensors, 10 Sauerbrey G. Verwendung von Schwingquarzen zur Wa¨gung du¨nner Schichten thus each latent variable is a linear combination of all the sensors: und zur Mikrowa¨gung. ZPhys1959; 155: 206. they search for characteristic smell prints of particular conditions 11 Kneepkens CM, Lepage G, Roy CC. The potential of the hydrocarbon breath test as in the urine headspace without distinguishing the constituent a measure of lipid peroxidation. Free Radic Biol Med 1994; 17: 127–160. VOCs. Over-fitting condition is avoided by the selection of the 12 Bajaj A, Miranda OR, Kim IB, Phillips RL, Jerry DJ, Bunz UH et al. Detection and correct number of LVs (see Table 1). Moreover, as can be observed differentiation of normal, cancerous, and metastatic cells using nanoparticle- in the loading plots in Figure 6, none of the sensors in the array polymer sensor arrays. Proc Natl Acad Sci USA. 2009; 106: 10912–10916. 13 Filipiak W, Sponring A, Filipiak A, Ager C, Schubert J, Miekisch W et al. TD-GC-MS shows negligible load values with respect to the other ones. analysis of volatile metabolites of human lung cancer and normal cells in vitro. Concerning sensors’ pertinence with respect to PCa diagnosis, the Cancer Epidemiol Biomarkers Prev 2010; 19: 182–195. possible discrimination of cancer is based on fingerprint 14 Williams H, Pembroke A. Sniffer dogs in the melanoma clinic? Lancet 1989; 1: 734. differentiation, thus the volatile biomarkers of PCa remain to be 15 Pickel D, Manucy GP, Walker DB, Hall SB, Walker JC. Evidence for canine olfactory identified, in order to use selective adsorbing material in sensor detection of melanoma. Appl Anim Behav Sci 2004; 89: 107–116. arrays. 16 Ehmann R, Boedeker E, Friedrich U, Sagert J, Dippon J, Friedel G et al. Canine scent Some limitations of our study should be mentioned. The detection in the diagnosis of lung cancer: revisiting a puzzling phenomenon. number of cases is small to permit definitive conclusions on the Eur Respir J 2012; 39: 669–676. diagnostic accuracy of the EN, although the preliminary data 17 Horvath G, Andersson H, Paulsson G. Characteristic odour in the blood reveals ovarian carcinoma. BMC Cancer 2010; 10: 643. appears at least promising. Only two cases of Gleason score 7 18 McCulloch M, Jezierski T, Broffman M, Hubbard A, Turner K, Janecki T. Diagnostic disease were included in the study, and none with Gleason accuracy of canine scent detection in early- and late-stage lung and breast scoreX8; consequently, the ability of the device in the diagnosis cancers. Integr Cancer Ther 2006; 5: 30–39. of high-grade PCa cannot be evaluated. However, it seems 19 Sonoda H, Kohnoe S, Yamazato T, Satoh Y, Morizono G, Shikata K et al. Colorectal promising that both cases of Gleason 7 were correctly classified by cancer screening with odour material by canine scent detection. Gut 2011; 60: the device as positive. The minimum number of positive cores/ 814–819. minimum percentage of core involvement that correlates to an 20 Willis CM, Church SM, Guest CM, Cook WA, McCarthy N, Bransbury AJ et al. accurate diagnosis were not addressed. However, in our prelimi- Olfactory detection of human bladder cancer by dogs: proof of principle study. nary analysis, there was not a significant difference between mean BMJ 2004; 329: 712. 21 Bijland LR, Bomers MK, Smulders YM. Smelling the diagnosis: a review on the use number of positive cores in correctly identified PCa vs misclassi- of scent in diagnosing disease. Neth J Med 2013; 71: 300–307. fied patients. Finally, any correlation between the EN results and 22 Gordon RT, Schatz CB, Myers LJ, Kosty M, Gonczy C, Kroener J et al. The use of canines the outcomes of the radical prostatectomy specimen was not in the detection of human cancers. J Altern Complement Med 2008; 14: 61–67. explored. 23 Leja MA, Liu H, Haick H. Breath testing: the future for digestive cancer detection. Table 3 suggests crucial areas for further future research on EN. Expert Rev Gastroenterol Hepatol 2013; 7: 389–391. 24 http://www.cnr.it/documenti/Istituzionali/Consuntivo/Report/Report2003_file/pdf/ 237-238Focus.pdf. CONCLUSIONS 25 Bajtarevic A, Ager C, Pienz M, Klieber M, Schwarz K, Ligor M et al. Noninvasive detection of lung cancer by analysis of exhaled breath. BMC Cancer 2009; 9: 348. To the best of our knowledge, this is the first demonstration of an 26 D’Amico A, Pennazza G, Santonico M, Martinelli E, Roscioni C, Galluccio G et al. olfactory imprinting of the initial part of the urine stream in An investigation on electronic nose diagnosis of lung cancer. Lung Cancer 2010; patients with PCa that was accurately revealed by an EN. The high 68: 170–176. specificity of the device, if confirmed in larger studies, may 27 Di Natale C, Macagnano A, Martinelli E, Paolesse R, D’Arcangelo G, Roscioni C et al. overcome the known specificity gap of the PSA serum test, and it Lung cancer identification by the analysis of breath by means of an array of may permit a more accurate selection of the PB candidates. non-selective gas sensors. Biosens Bioelectron 2003; 18: 1209–1218. 28 Machado RF, Laskowski D, Deffenderfer O, Burch T, Zheng S, Mazzone PJ et al. Detection of lung cancer by sensor array analyses of exhaled breath. Am J Respir CONFLICT OF INTEREST Crit Care Med 2005; 171: 1286–1291. 29 Barash O, Peled N, Hirsch FR, Haick H. Sniffing the unique "odor print" of non- The authors declare no conflict of interest. small-cell lung cancer with gold nanoparticles. Small 2009; 5: 2618–2624. 30 Peled N, Hakim M, Bunn Jr PA, Miller YE, Kennedy TC, Mattei J et al. Non-invasive breath analysis of pulmonary nodules. J Thorac Oncol 2012; 7: 1528–1533. REFERENCES 31 Xu ZQ, Broza YY, Ionsecu R, Tisch U, Ding L, Liu H et al. A nanomaterial-based 1 Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun MJ. Cancer statistics, 2009. CA Cancer J breath test for distinguishing gastric cancer from benign gastric conditions. Clin 2009; 59: 225–249. Br J Cancer 2013; 108: 941–950. 2 Schro¨der FH. PSA screening—a review of recent studies. Eur J Cancer 2009; 32 Horvath G, Chilo J, Lindblad T. Different volatile signals emitted by human ovarian 45(Suppl 1): 402–404. carcinoma and healthy tissue. Future Oncol 2010; 6: 1043–1049. 3 Stamey TA, Caldwell M, McNeal JE, Nolley R, Hemenez M, Downs J. The prostate 33 Peng G, Hakim M, Broza YY, Billan S, Abdah-Bortnyak R, Kuten A et al. Detection of specific antigen era in the United States is over for prostate cancer: what lung, breast, colorectal, and prostate cancers from exhaled breath using a single happened in the last 20 years? J Urol 2004; 172: 1297–1301. array of nanosensors. Br J Cancer 2010; 103: 542–551. 4 Tuma RS. New tests for prostate cancer may be nearing the clinic. J Natl Cancer 34 Phillips M, Cataneo RN, Saunders C, Hope P, Schmitt P, Wai J. Volatile biomarkers Inst 2010; 102: 752–754. in the breath of women with breast cancer. J Breath Res 2010; 4: 026003. 5 Matsumura K, Opiekun M, Oka H, Vachani A, Albelda SM, Yamazaki K et al. Urinary 35 D’Amico A, Bono R, Pennazza G, Santonico M, Mantini G, Bernabei M et al. volatile compounds as biomarkers for lung cancer: a proof of principle study Identification of melanoma with a gas sensor array. Skin Res Technol 2008; 14: using odor signatures in mouse models of lung cancer. PLoS One 2010; 5: e8819. 226–236. 6 Cornu JN, Cancel-Tassin G, Ondet V, Girardet C, Cussenot O. Olfactory detection of 36 Bernabei M, Pennazza G, Santonico M, Corsi C, Roscioni C, Paolesse R et al. prostate cancer by dogs sniffing urine: a step forward in early diagnosis. Eur Urol A preliminary study on the possibility to diagnose urinary tract cancers by an 2011; 59: 197–201. electronic nose. Sens Actuators B Chem 2008; 131: 1–4.

& 2014 Macmillan Publishers Limited Prostate Cancer and Prostatic Disease (2014), 206 – 211