br were performed using the open
were performed using the open-source statistical computing soft-ware R.23
The study analyzed VOCs in urine samples collected from patients to develop the urine metabolome-based PCa diagnosis model. All VOCs were identified using an existing National Insti-tute of Standards and Technology library, and significant VOCs were selected based on their occurrence and relative quantity in the urine. One mL of urine sample was extracted by stir bar sorptive
extraction, and the VOCs were then analyzed by GC/MS. The quantity of each VOC was normalized to the internal standard, Mirex, by taking the ratio between the signal of the AR7 and that of Mirex.
A total of 9144 potential VOCs were detected in urine from 108 patients (55 PCa-positive and 53 PCa controls). Using the Wil-coxon test, 254 VOCs were found to be positively related to PCa, and 282 VOCs were negatively associated at P < .05 (Figure 1A). To avoid missing important predictors for PCa prevalence, a relatively large threshold (with P < .20) was applied to screen variables for further development of the regression model, resulting in a total of 850 potential VOCs. After l1 regularization (ie, lasso regression), the final logistic model contained 11 VOCs (listed in Table 2), and a heat map was generated to show their differentiation in PCa-positive and -negative patients (Figure 1B). On the basis of predicted probabilities from the final model obtained via jackknife cross-validation, the area under the ROC curve (AUC) is 0.92, with sensitivity of 0.96 and a specificity of 0.80, respectively, as shown in Figure 2A. As a com-parison, we also built a logistic model with PSA only, which yielded an AUC of 0.54. The model resulted in a sensitivity of 0.60 and a specificity of 0.42, respectively, when applied to the test data (Figure 2B). Testing the performance of the PCa diagnostic model using an external cohort of 75 patients (53 patients with PCa and 22 PCa-negative patients) yielded a ROC curve of 0.86 (Figure 3), with an 87% sensitivity and a 77% specificity. r> Discussion
There is significant interest in PCa biomarker development as evidenced by the number of biomarkers currently available for disease diagnosis, risk stratification, prognosis, and response to treatment.24 In this study, we developed a urinary VOCs based model for PCa diagnosis. The analytical method allowed for an easy, fast, and efficient analysis of VOCs without tedious sample prepa-ration. The solventless sample preparation technique preserved the sample integrity and permitted effective analyses for processing large number of samples, which is an important factor for clinical translatability. Unlike gas-sensors (eg, E-nose), which are based on signal indications,16 the GC-MS method identified relevant com-pounds that can be further validated or interrogated in future
Urinary VOCs for Prostate Cancer Diagnosis
Figure 1 A, The Heat Map of Significant VOCs by the Wilcoxon Test (P < .05) in PCa Versus PCa-negative Samples. The Correlation Between VOCs and Patients Ranges From Low (Red) to High (Blue); B, the Heat Map of 11 Selected VOCs in the Urinary VOC-based PCa Diagnostic Model. The Values of Those Selected VOCs in Patients Show a Strong Pattern in Distinguishing PCa-Positive and -negative Patients
Abbreviations: PCa ¼ prostate cancer; VOCs ¼ volatile organic compounds.
metabolomics and physiological studies. Using the internal stan-dard, Mirex, facilitated the semi-quantitative determination of VOCs. The expression of VOCs as continuous data rather than as binary provides a more disease-specific representation of metabolite concentration (ie, increase or decrease) in vivo, and improves the subsequent model development.
For the urinary VOC-based PCa diagnostic model, 11 VOCs were selected into the final logistic regression. The model was validated and produced an AUC of 0.92 (Figure 2A), suggesting the potential of urinary VOCs for PCa diagnosis. Although other studies have evaluated various urinary analytic techniques for PCa detection, the study by Khalid et al represents the best study to date,
utilizing urinary VOCs for this purpose. Their study was comprised of 59 patients with PCa and 43 controls. Following urinary VOC analysis of PCa-positive and -negative specimens, they manually selected 4 VOCs (2,6-dimethyl-7-octen-2-ol, pentanal, 3-octanone, and 2-octanone) for their PCa detection model. The accuracy of this model was 63% to 65% but improved to 74% and 65% when combined with PSA using their random forest and linear discrimi-nant analysis methods, respectively.17 In their study, PSA alone was marginally better than the 4-VOC model, which calls into question the clinical utility of this set of biomarkers. By comparison, in our study, there was a significant difference between the performance of PSA (AUC of 0.54) and the diagnostic model (AUC of 0.92).