STAT 115: SECTION 11 Research Example Evaluation of a urine test for detection of bladder cancer Urine Test for Bladder Cancer THE JOURNAL OF UROLOGY® Vol. 188, 741-747, September 2012 Bladder Cancer http://www.urologyhealth.org/urology/articles/images/anatomy_bladdercancer.jpg Bladder Cancer www.cancer.gov/cancertopics/types/bladder Detection of bladder cancer Standard practice for detection of bladder cancer includes: Cystoscopy • a cystoscope hollow tube with a lens at the end is passed through the ureter and into the bladder • Can be flexible or rigid • Used to visualise the inside the bladder to see if there are any abnormal areas • Abnormal areas are biopsied – a small piece of tissue is removed and sent to the pathologist to determine if there is cancer Cytology: • Cells in a urine sample are looked at to see if there are any abnormal cells that might have been washed off the bladder wall. Other urine tests: eg Nuclear Matrix Protein 22 (NMP22™) Performance of a diagnostic test Sensitivity measures how good the test is as identifying people with disease measured on a group of people with the disease Specificity measures how good the test is in ruling out disease in people who don’t have disease measured on a group of people without the disease Continuous diagnostic tests Test which gives a continuous measure Disease Disease free Increased specificity Increased sensitivity Test negative Cut-off Test positive Detection of bladder cancer Cystoscopy in combination with biopsy and histopathological diagnosis: Sensitivity of over 90% Specificity of 100% Although examination with flexible instruments is considered to be a rapid routine procedure, cystoscopy is invasive and associated with a cost. Urinary cytology: Sensitivity 34% Specifcity 99% Urine cytology is also relatively expensive and requires an experienced pathologist and meticulous sample collection, storage and preparation. Regardless, it is still in routine use as an adjunct to cystoscopy. Development of urine test: Phase I Clin Cancer Res 2008;14(3) Original development of the test • identified 4 RNA markers CDC2, MDK, IGFBP5, and HOXA13, which distinguished TCC and non-cancer tissue • used statistical methods to combine these continuous marker variables into a single score (uRNA-D®) which discriminated between bladder cancer and non-bladder cancer samples Continuous diagnostic tests Test which gives a continuous measure Disease Disease free Increased specificity Increased sensitivity Test negative Cut-off Test positive Development of urine test: Phase I Urine samples were obtained from • 75 bladder cancer patients (with Transitional Cell Carcinoma) • 77 patients with nonmalignant diseases or other cancers, in whom bladder tumors were excluded by flexible cystoscopy. • The uRNA-D® test consists of qRT-PCR reactions for the four RNA markers MDK, CDC2, HOXA13 and IGFBP5. Each of these markers is up-regulated in transitional cell carcinoma (TCC). • These early studies found for cut off fixed at a specificity of 85% • Sensitivity = 48/75 = 60% • Stage Ta; sensitivity = 48%, • Stage T1; sensitivity = 90%, • Stage >T1: Sensitivity= 100% Development of diagnostic test: Phase II THE JOURNAL OF UROLOGY® Vol. 188, 741-747, September 2012 Development of diagnostic test: Phase II Primary Objective: To determine the characteristics (sensitivity, specificity, area under the ROC curve, positive and negative predictive values) of the uRNA-D® test for the detection of TCC in patients with a recent history of gross haematuria. Design: • Cohort study • Participants: patients with a recent history of gross haematuria, who are undergoing investigation of haematuria (by cystoscopy) for possible urological cancer. • Eligible consenting patients provided a freshly voided mid-stream urine sample prior to cystoscopy, for uRNA®, the NMP22™ tests and urine cytology analysis. The presence of urinary tract TCC was determined by biopsy and histopathological examination within a 3 months Results of Phase II A number of the 594 patients registered excluded from analysis • inadequate determination of TCC status (n=18): • lack of availability of urine test (n=13): Figure 5. Test results for uRNA-D® by bladder cancer status and by area Figure 6. Test results for NMP22 ELISA by bladder cancer status and area. Results of Phase II Using data from New Zealand and Australia only Sensitivity (at specificity of 85%): Number with bladder cancer = 66 Number with a positive uRNA-D test = 41 Sensitivity = 41/66= 0.621 or 62.1% Figure 7. Comparison of ROC curves for uRNA-D® and NMP22 ELISA for New Zealand and Australia . Figure 16. ROC curve for Cxbladder® (a refinement of the uRNA-D test based on the cohort study data) . Confidence interval calculated using an exact method, so is slightly wider than the method taught in this class, which was