Optimizing the early diagnosis process of urinary bladder cancer. On standardized care pathway, computed tomography, artificial intelligence, and a newly developed urine-based cancer marker

Abstract

The thesis contains of five articles in three different subprojects. The overall aim of this thesis is to investigate the development of the investigation of macroscopic hematuria over time, the effects of this development on urinary bladder cancer (UBC) characteristics and possible improvements to the investigation process using auxiliary tools such as computed tomography urography (CTU), artificial intelligence (AI) interpretation of CTU and urine-based cancer marker tests. We have therefore studied the impact of early detection of UBC on its characteristics in the first two articles (I and II) which were retrospective, observational, cohort studies (2010-2019). We could show that, following the implementation of the standardized clinical pathway (SCP) in Sweden for patients with macroscopic hematuria, the median time to diagnosis for UBC decreased nationally from 37 to 27 days and in NU Hospital Group from 29 to 12 days. The percentage of cT2-4 tumors decreased in the NU Hospital Group from 26% to 20% (p=0.035) during SCP. To enhance the diagnostic process, we propose stricter adherence to SCP and extending the 13-day window for transurethral resection of bladder tumors (TURBT), particularly in less urgent cases, to prioritize severe cases with treatable diseases for prompt assessment. We also evaluated the potential of CTU as a single investigation method as well as AI interpretation of CTU to improve the investigation of patients with suspected UBC in articles III and IV. Article III is a retrospective diagnostic accuracy study (2016-2019), and article IV is an observational case-control study (2016-2022). We showed that CTU has a high accuracy in detection of UBC with a false negative rate of 0.07 (95% CI 0.04-0.12), and a negative predictive value (NPV) of 0.99 (95% CI 0.92-1.0). As a result, our studies demonstrate that CTU can effectively rule out UBC, suggesting that forgoing cystoscopy may be a reasonable approach in 57% of patients. Our newly developed AI-based model has a sensitivity of 0.83 (95% CI, 0.76-0.89), and NPV of 0.97 (95% CI 0.95-0.98). This AI-based image analysis model may assist radiologists in the initial assessment of CTUs for patients with macroscopic hematuria. Finally, we have evaluated a new urine test with an mRNA panel (GeneXpert BC) to detect or exclude UBC in article V which was a multicenter nested case-control diagnostic study (2020-2022). In this study, involving 273 subjects (case group, n=151, and control group, n=122), the sensitivity of GeneXpert BC was 0.94 (95% CI 0.89-0.97), and NPV was 0.99 (95% CI 0.97-1.00). Consequently, GeneXpert BC was shown as a reliable triage test in ruling out UBC, suggesting that 44% of our patients might have been exempted from routine primary investigation of macroscopic hematuria which might potentially save healthcare resources and spare patients the discomfort of unnecessary examinations.