Evaluating the predictive value of multiple rounds of faecal haemoglobin measurement for the prediction of advanced neoplasia and colorectal cancer: an analysis of data from the Dutch national colorectal cancer screening programme.
Faecal haemoglobin (f-Hb) concentrations from two screening rounds improve colorectal cancer (CRC) risk prediction, but the value of longer longitudinal data remains uncertain. This study aimed to assess whether incorporating f-Hb data from three prior screening rounds improves risk stratification for CRC and advanced neoplasia (AN).
Using data from four rounds of the Dutch national CRC screening programme (1 January 2014 and 31 December 2022), prognostic models combining age, sex, and f-Hb concentrations were developed to predict CRC and AN. Multiple representations of longitudinal f-Hb were evaluated. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration, and information criteria. Risk reclassification was quantified using net reclassification improvement (NRI) and integrated discrimination improvement (IDI). The policy impact of risk-based referral, expressed as additional colonoscopies, was examined. External validation used data from a Dutch faecal immunochemical test pilot study.
Among 1 171 070 participants included, 2357 (0.2%) had CRC and 14 543 (1.2%) had AN. Models incorporating categorised f-Hb values from three prior rounds, age, and sex achieved the highest discrimination (AUC: AN 0.76 (0.75-0.77); CRC 0.73 (0.72-0.74)) and significantly improved reclassification (NRI and IDI, p<0.001). Participants in the highest 15% predicted risk had a 6.8-fold (95% CI 6.4 to 6.9) increased risk of AN and a 5.9-fold (95% CI 5.4 to 6.6) increased risk of CRC. Risk-based referral enabled earlier detection with minimal additional colonoscopy for both CRC and AN.
Including f-Hb data from three prior screening rounds improves CRC and AN risk prediction. Targeted referral of the highest risk individuals enables earlier detection with limited impact on colonoscopy capacity, supporting personalised CRC screening strategies.
Using data from four rounds of the Dutch national CRC screening programme (1 January 2014 and 31 December 2022), prognostic models combining age, sex, and f-Hb concentrations were developed to predict CRC and AN. Multiple representations of longitudinal f-Hb were evaluated. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration, and information criteria. Risk reclassification was quantified using net reclassification improvement (NRI) and integrated discrimination improvement (IDI). The policy impact of risk-based referral, expressed as additional colonoscopies, was examined. External validation used data from a Dutch faecal immunochemical test pilot study.
Among 1 171 070 participants included, 2357 (0.2%) had CRC and 14 543 (1.2%) had AN. Models incorporating categorised f-Hb values from three prior rounds, age, and sex achieved the highest discrimination (AUC: AN 0.76 (0.75-0.77); CRC 0.73 (0.72-0.74)) and significantly improved reclassification (NRI and IDI, p<0.001). Participants in the highest 15% predicted risk had a 6.8-fold (95% CI 6.4 to 6.9) increased risk of AN and a 5.9-fold (95% CI 5.4 to 6.6) increased risk of CRC. Risk-based referral enabled earlier detection with minimal additional colonoscopy for both CRC and AN.
Including f-Hb data from three prior screening rounds improves CRC and AN risk prediction. Targeted referral of the highest risk individuals enables earlier detection with limited impact on colonoscopy capacity, supporting personalised CRC screening strategies.
Authors
Mwangi Mwangi, Brück Brück, Katsara Katsara, Toes-Zoutendijk Toes-Zoutendijk, Vuuren Vuuren, Thelen Thelen, Nagetaal Nagetaal, Dekker Dekker, Spaander Spaander, Leerdam Leerdam, Lansdorp Vogelaar Lansdorp Vogelaar
View on Pubmed