Application of Urinary Metabolites for Cancer Detection

Qin Gao, University of Texas at El Paso

Abstract

Prostate cancer (PCa) is the 3rd most common cause of male cancer mortality in the US. Early diagnosis and treatment of PCa will improve the quality of care and reduce mortality. The prostate specific antigen (PSA) is commonly used in the current PCa screening, but its limitation has resulted in an intense search for more reliable biomarkers. Studies showed that dogs could differentiate PCa patients from negative control by sniffing their urine. As the odor profiles are generated by volatile organic compounds (VOCs), the finding suggests that particular VOCs could be linked to PCa, PCa risk levels and other cancers. Therefore the overarching goal of study was to develop a reliable, quick and patient-friendly diagnostic method for PCa screening to replace the current PSA testing and avoid unnecessary biopsy. The objectives of this study were 1) to develop and validate a urinary VOCs based model for PCa diagnosis; 2) to study the performance of urinary VOCs in differentiating high/intermediate PCa patients from low risk PCa patients; 3) to investigate the urinary VOCs for the early screening of other urological cancer, such as clear cell renal cell carcinoma (ccRCC); and 4) to evaluate the specificity of urinary VOCs models in PCa and ccRCC detection.In the study of the PCa urinary VOCs profile, a VOCs based PCa model was developed and validated. Urine samples from 55 and 53 biopsy proven PCa positive and negative patients respectively were obtained for diagnostic model development. Urinary metabolites were analyzed by Stir Bar Sorptive Extraction coupled with Gas Chromatography-Mass Spectrometry. A PCa diagnosis model was developed and validated using innovative statistical machine-learning techniques. The analysis resulted in 254 and 282 VOCs for their significant association (p < 0.05) with either PCa positive or negative samples respectively. Regularized logistic regression analysis and the Firth method were then applied to predict PCa prevalence, resulting in a final model that contained 11 VOCs. Under cross-validation, the area under the receiver operating characteristic curve (AUC) for the final model was 0.92 (sensitivity: 0.96; specificity: 0.80). Further evaluation of the developed model using a testing cohort yielded an AUC of 0.86. As a comparison, the PSA-based diagnosis model of the same cohort of patients only rendered an AUC of 0.54.Then, since the performance of VOCs was proved to be able to strongly discriminate PCa patients from controls, we hypothesized that urinary VOCs could be used to develop urinary VOC PCa prognostic models for cancer risk assessment. PCa is a heterogeneous disease ranging from indolent to life threatening stages. Another VOCs based PCa risk assessment model was developed through comparing high/intermediate risk patients with low risk patients. Based on the D’Amico risk scores system, these PCa patients were divided into two groups: 55 in low-risk group (indolent PCa, GS = 6, PSA < 10) and 34 in high/intermediate-risk group (clinically significant PCa, GS = 6 and PSA ≥ 10, or GS > 6 with any PSA values). Urine samples from 89 men who presented for transrectal ultrasound guided prostate biopsy for an elevated PSA or abnormal digital rectal exam were collected. Using the Wilcoxon rank sum test, 23 VOCs were found to be positively related to high/intermediate−risk PCa and 44 VOCs negatively associated. Regularized logistic regression analysis and the Firth method were then applied to predict PCa risk level, resulting in a final model that contained 11 VOCs. Under cross-validation, the area under the receiver operating characteristic curve (AUC) for the final model was 0.86 (sensitivity: 0.85; specificity: 0.80), which indicates a highly promising discrimination power of urinary VOCs in PCa high/intermediate risk assessment.To test the performance of VOCs in other genitourinary cancers, we investigated the urinary VOCs profile of clear cell renal cell carcinoma (ccRCC the main type of renal cell carcinoma, RCC). The fast and reliable early screen of RCC enables better outcome and predication in patients. However, there is no recommended screening tests for RCC available clinically. A total of 108 urine samples were obtained from 71 patients who were undergoing partial or radical nephrectomy and 37 patients ccRCC negative based the imaged renal mass to identify the specific VOCs in urine for ccRCC screening. The VOCs based ccRCC diagnostic model was developed through the logistic regression in training set (57 ccRCC vs 31 controls) and validated in another testing set group (14 RCC vs 6 controls). A total of 8,266 VOCs were found in the urine samples of training set. Using Wilcoxon Rank Sum Test in screening their bivariate association with ccRCC, 79 VOCs were found to be related to urine samples of ccRCC patients while 91 VOCs corresponding to RCC negative controls with statistical significance at p = 0.05. After further selection with ℓ1 regularization, 15 VOCs were included in the RCC diagnostic logistic model. On the basis of predicted probabilities from the model via cross-validation, the area under the receiver operating characteristic curve (AUC) was found to be 0.87 and the sensitivity and specificity were 93% and 77% respectively. The VOCs based RCC diagnostic model were then validated in testing group. The results showed a promising diagnostic power of this VOCs model in ccRCC screening with AUC of 0.81, with 86% sensitivity and 83% specificity respectively.Moreover, we cross examined the performance of the two PCa and ccRCC VOCs models. Four cohorts from the previous studies were involved in this analysis. The four groups contained (1) 55 PCa positive patients, PCa (+), (2) 53 PCa negative patients, PCa (−), (3) 55 ccRCC positive patients, ccRCC (+), and (4) 31 RCC negative patients, ccRCC (−). For the PCa model was cross examined between PCa (+) and the rest of the three groups combined, and the result showed an AUC of 0.834 with confidence interval 0.779 to 0.889. Then, the ccRCC model was validated through comparing ccRCC (+) with the rest of the three groups combined. The validation rendered an AUC of 0.779 with confidence interval 0.707 to 0.851. The cross evaluation verified the discrimination power of those 11 VOCs based PCa model and 15 VOCs based ccRCC model even in more complicated patients.The investigations among PCa and ccRCC demonstrate and validate the clinical utility of a non-invasive urinary VOCs based diagnostic model in PCa and ccRCC screening. The VOCs based diagnostic model has the substantial potency and clinical value in PCa and RCC screening, and the analytical method was fast and highly translatable.

Subject Area

Analytical chemistry|Statistics|Oncology

Recommended Citation

Gao, Qin, "Application of Urinary Metabolites for Cancer Detection" (2019). ETD Collection for University of Texas, El Paso. AAI13883354.
https://scholarworks.utep.edu/dissertations/AAI13883354

Share

COinS