Development and validation of a haematuria cancer risk score to identify patients at risk of harbouring cancer

Abstract

Background: A lack of consensus exists amongst national guidelines regarding who should be investigated for haematuria. Type of haematuria and age-specific thresholds are frequently used to guide referral for the investigation of haematuria. Objectives: To develop and externally validate the haematuria cancer risk score (HCRS) to improve patient selection for the investigation of haematuria. Methods: Development cohort comprise of 3539 prospectively recruited patients recruited at 40 UK hospitals (DETECT 1; ClinicalTrials.gov: NCT02676180) and validation cohort comprise of 656 Swiss patients. All patients were aged >18 years and referred to hospital for the evaluation of visible and nonvisible haematuria. Sensitivity and specificity of the HCRS in the validation cohort were derived from a cut-off identified from the discovery cohort. Results: Patient age, gender, type of haematuria and smoking history were used to develop the HCRS. HCRS validation achieves good discrimination (AUC 0.835; 95% CI: 0.789–0.880) and calibration (calibration slope = 1.215) with no significant overfitting (P = 0.151). The HCRS detected 11.4% (n = 8) more cancers which would be missed by UK National Institute for Health and Clinical Excellence guidelines. The American Urological Association guidelines would identify all cancers with a specificity of 12.6% compared to 30.5% achieved by the HCRS. All patients with upper tract cancers would have been identified. Conclusion: The HCRS offers good discriminatory accuracy which is superior to existing guidelines. The simplicity of the model would facilitate adoption and improve patient and physician decision-making.