Elucidating bleeding episodes and the clinical impact of bleeding

The overall purpose of this PhD project was to elucidate bleeding history using different methodsand determine the clinical impact of bleeding. We aimed to (1) develop a Danish version of thepatient administrated bleeding assessment tool (self-BAT), (2) determine the cut-off betweenabnormal and normal overall bleeding tendency, along with the prevalence of bleeding in Denmark,(3) determine the prevalence of bleeding from any anatomical location in individuals with bladdercancer using an artificial intelligence (AI) model, and (4) determine the prevalence of prostatecancer in individuals with blood in the urine (haematuria).

Study one: Self-BAT is a validated patient-reported outcome measure for obtaining a detailed bleeding history. However, no Danish version of self-BAT exists. We therefore developed a Danish version of self-BAT and compared it with the internationally recognized bleeding assessment tool from international society of thrombosis and haemostasis (ISTH-BAT) for 106 individuals.

Study two: Large cohort studies on self-BAT scores of individuals considered healthy is lacking, as well as knowledge on the prevalence of bleeding in Denmark. We invited blood donors and 33.800 individual’s representative of the Danish population to complete the self-BAT. We found cut-offs for abnormal total self-BAT score to be higher than the existing ones, and we established the cut-offs for abnormal total self-BAT score in children. We found that bleeding episodes were very common in both Danish children and adults, with epistaxis, menorrhagia, cutaneous bleeding, and bleeding from minor wounds being the most common anatomical locations of bleeding.

Study three: Besides questionnaires like the self-BAT, the bleeding history of patients can be obtained from the electronic health record (EHR). Unfortunately, it is a time-consuming process to
manually uncover all bleeding episodes in the EHR, with risk of overlooking important information, why new methods for identifying bleeding episodes in the EHR is needed. Therefore, an AI model for identification of sentences indicative of bleeding in EHR text was developed by engineers in close collaboration with our team. Urological bleeding is known to be associated with bladder cancer, but knowledge on risk factors for bleeding and knowledge on the overall bleeding tendency in individuals with bladder cancer is lacking. We applied the AI model on the EHR from one year before bladder cancer diagnosis to two years after bladder cancer diagnosis in 162 individuals. We found that age, gender, and smoking at time of diagnosis did not increase the risk for developing bleeding after bladder cancer diagnosis. We found that 26% of individuals with bladder cancer had at least one non-urological bleeding before diagnosis. However, only 1.8% experienced nonurological bleeding without also experiencing urological bleeding before diagnosis, why solitary non-urological bleeding cannot be used to identify individuals with undiagnosed bladder cancer. Furthermore, we found that bleeding risk increased with tumor burden, but metastasis did not seem to increase the risk of bleeding.

Study four: Haematuria can arise due to local cancer effects and many studies on haematuria focus on bladder or kidney cancer. However, haematuria can also arise from prostate cancer, but no systematic review or meta-analysis had been performed on the prevalence of prostate cancer in men presenting with haematuria. We scrutinized 4252 titles and available abstracts on haematuria
and prostate cancer, of assessed 350 records in full text, of which 65 studies were included in a systematic review and meta-analysis. We found that 3% of men attending work-up for haematuria had prostate cancer, indicating that men with haematuria should be examined for prostate cancer.

In conclusion, this PhD project moved the field of bleeding research forward by (1) developing a Danish questionnaire for elucidating bleeding episodes, (2) establishing gender specific cut-offs for abnormal and normal overall bleeding tendency in adults and children, while establishing that bleeding is prevalent in Denmark, especially epistaxis, menorrhagia, cutaneous bleeding, and bleeding from minor wounds, (3) using an AI model to establish that bleeding is common in individuals with bladder cancer but sole non-urological bleeding does not predict bladder cancer, and (4) establishing that men with hematuria should undergo tests for prostate cancer as the prevalence of prostate cancer in men with hematuria is around 3%.

Next step is testing the Danish self-BAT in different clinical settings and use the AI model to determine the clinical impact of bleeding in individuals with prostate cancer.